research paper topics biometrics

A Systematic Review on Physiological-Based Biometric Recognition Systems: Current and Future Trends

• Review article
• Published: 23 February 2021
• Volume 28 , pages 4917–4960, ( 2021 )

Cite this article

• Kashif Shaheed   ORCID: orcid.org/0000-0002-7399-6211 1 ,
• Aihua Mao 1 ,
• Imran Qureshi 2 ,
• Munish Kumar 3 ,
• Qaisar Abbas 4 ,
• Inam Ullah 5 &
• Xingming Zhang 1  

1868 Accesses

24 Citations

1 Altmetric

Explore all metrics

Biometric deals with the verification and identification of a person based on behavioural and physiological traits. This article presents recent advances in physiological-based biometric multimodalities, where we focused on finger vein, palm vein, fingerprint, face, lips, iris, and retina-based processing methods. The authors also evaluated the architecture, operational mode, and performance metrics of biometric technology. In this article, the authors summarize and study various traditional and deep learning-based physiological-based biometric modalities. An extensive review of biometric steps of multiple modalities by using different levels such as preprocessing, feature extraction, and classification, are presented in detail. Challenges and future trends of existing conventional and deep learning approaches are explained in detail to help the researcher. Moreover, traditional and deep learning methods of various physiological-based biometric systems are roughly analyzed to evaluate them. The comparison result and discussion section of this article indicate that there is still a need to develop a robust physiological-based method to advance and improve the performance of the biometric system.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Price includes VAT (Russian Federation)

Instant access to the full article PDF.

Rent this article via DeepDyve

Institutional subscriptions

Similar content being viewed by others

An Overview of Deep Learning Techniques for Biometric Systems

State-of-the-Art Multi-trait Based Biometric Systems: Advantages and Drawbacks

Biometrics recognition using deep learning: a survey

Abbas EI, Mieee MES (2017) Face recognition rate using different classifier methods based on PCA, pp 37–40

Abrishami-Moghaddam H, Farzin H, Moin MS (2008) A novel retinal identification system. Eurasip J Adv Signal Process. https://doi.org/10.1155/2008/280635

Article   MATH   Google Scholar  

Adjimi A, Hacine-Gharbi A, Ravier P, Mostefai M (2017) Extraction and selection of binarised statistical image features for fingerprint recognition. Int J Biometrics 9(1):67–80. https://doi.org/10.1504/IJBM.2017.084133

Article   Google Scholar  

Aglio-caballero A, Ríos-Sánchez B, Sánchez-Ávila C, Giles MJMD (2017) Analysis of local binary patterns and uniform local binary patterns for palm vein biometric recognition. In: 2017 international carnahan conference on security technology (ICCST). IEEE, pp 1–6

Ahmadi N, Akbarizadeh G (2018) Hybrid robust iris recognition approach using iris image pre-processing, two-dimensional gabor features and multi-layer perceptron neural network/PSO. IET Biometrics 7(2):153–162. https://doi.org/10.1049/iet-bmt.2017.0041

Ahmadi N, Akbarizadeh G (2018) Iris tissue recognition based on GLDM feature extraction and hybrid MLPNN-ICA classifier. Neural Comput Appl. https://doi.org/10.1007/s00521-018-3754-0

Ahmadi N, Nilashi M, Samad S, Rashid TA, Ahmadi H (2019) An intelligent method for iris recognition using supervised machine learning techniques. Opt Laser Technol 120(2018):105701. https://doi.org/10.1016/j.optlastec.2019.105701

Ahmed T, Sarma M (2018) An advanced fingerprint matching using minutiae-based indirect local features. Multimed Tools Appl 77(15):19931–19950. https://doi.org/10.1007/s11042-017-5444-9

Aleem S, Sheng B, Li P, Yang P, Feng DD (2019) Fast and accurate retinal identification system: using retinal blood vasculature landmarks. IEEE Trans Ind Inf 15(7):4099–4110. https://doi.org/10.1109/TII.2018.2881343

Ali YH, Razuqi ZN (2017) Palm vein recognition based on centerline. Iraqi J Sci 58(2):726–734

Google Scholar  

Ananth C (2017) Iris recognition using active contours. SSRN Electron J 2(1):27–32. https://doi.org/10.2139/ssrn.3028575

Arsalan M, Hong HG, Naqvi RA, Lee MB, Kim MC, Kim DS et al (2017) Deep learning-based iris segmentation for iris recognition in visible light environment. Symmetry. https://doi.org/10.3390/sym9110263

Barpanda SS, Sa PK, Marques O, Majhi B, Bakshi S (2018) Iris recognition with tunable filter bank based feature. Multimed Tools Appl 77(6):7637–7674. https://doi.org/10.1007/s11042-017-4668-z

Benalcazar DP, Perez CA, Bastias D, Bowyer KW (2019) Iris recognition: comparing visible-light lateral and frontal illumination to NIR frontal illumination. In: Proceedings—2019 IEEE winter conference on applications of computer vision, WACV 2019, pp 867–876. https://doi.org/10.1109/WACV.2019.00097

Bharadwaj S, Vatsa M, Singh R (2014) Biometric quality: a review of fingerprint, iris, and face. Eurasip J Image Video Process 2014(1):1–28. https://doi.org/10.1186/1687-5281-2014-34

Bhardwaj I, Londhe ND, Kopparapu SK (2017) A spoof resistant multibiometric system based on the physiological and behavioral characteristics of fingerprint. Pattern Recognit 62:214–224. https://doi.org/10.1016/j.patcog.2016.09.003

Bhavani M (2013) Human identification using finger images. Int J Comput Trends Technol 4(1):258–263

MathSciNet   Google Scholar  

Bhukya S (2019) A hybrid biometric identification and authentication system with retinal verification using AWN classifier for enhancing security. https://doi.org/10.1007/978-981-13-1580-0

Borra SR, Reddy GJ, Reddy ES (2016) A broad survey on fingerprint recognition systems. In: Proceedings of the 2016 IEEE international conference on wireless communications, signal processing and networking, WiSPNET 2016, pp 1428–1434. https://doi.org/10.1109/WiSPNET.2016.7566372

Boubchir L, Aberni Y, Daachi B (2018) Competitive coding scheme based on 2D log-gabor filter for palm vein recognition. In: 2018 NASA/ESA conference on adaptive hardware and systems (AHS), vol (i), pp 152–155

Buciu I, Gacsadi A (2016) Biometrics systems and technologies: a survey. Int J Comput Commun Control 11(3):315–330. https://doi.org/10.15837/ijccc.2016.3.2556

Cancian P, Di Donato GW, Rana V, Santambrogio MD, Elettronica D, Bioingegneria I, Milano P (2017) An embedded gabor-based palm vein recognition system, pp 405–408

Cao K, Jain AK (2018) Fingerprint indexing and matching: an integrated approach. In: IEEE international joint conference on biometrics, pp 437–445. https://doi.org/10.1109/BTAS.2017.8272728

Cao K, Jain AK (2019) Automated latent fingerprint recognition. IEEE Trans Pattern Anal Mach Intell 41(4):788–800. https://doi.org/10.1109/TPAMI.2018.2818162

Chauhan S, Arora AS, Kaul A (2010) A survey of emerging biometric modalities. Procedia Comput Sci 2:213–218. https://doi.org/10.1016/j.procs.2010.11.027

Chen Z, Huang W, Lv Z (2015) Towards a face recognition method based on uncorrelated discriminant sparse preserving projection. https://doi.org/10.1007/s11042-015-2882-0

Choraś M (2007) Human lips recognition. Adv Soft Comput 45:838–843. https://doi.org/10.1007/978-3-540-75175-5_104

Choraś M (2010) The lip as a biometric. Pattern Anal Appl 13(1):105–112. https://doi.org/10.1007/s10044-008-0144-8

Article   MathSciNet   Google Scholar  

Chugh T, Cao K, Jain AK (2018) Fingerprint spoof detection using minutiae-based local patches. In: IEEE international joint conference on biometrics, IJCB 2017, 2018 January, pp 581–589. https://doi.org/10.1109/BTAS.2017.8272745

Czajka A, Bowyer KW (2018) Presentation attack survey. ACM Comput Surv 51(4):1–35. https://doi.org/10.1145/3232849

Czajka A, Bowyer KW (2018) Presentation attack detection for iris recognition: an assessment of the state-of-the-art. ACM Comput Surv 51(4):86

Czajka A, Moreira D, Bowyer KW, Flynn PJ (2019) Domain-specific human-inspired binarized statistical image features for Iris recognition. In: Proceedings of IEEE winter conference on applications of computer vision, WACV 2019, pp 959–967. https://doi.org/10.1109/WACV.2019.00107

Dargan S, Kumar M (2020) A comprehensive survey on the biometric recognition systems based on physiological and behavioral modalities. Expert Syst Appl 143:113114. https://doi.org/10.1016/j.eswa.2019.113114

Darlow LN, Rosman B (2018) Fingerprint minutiae extraction using deep learning. In: IEEE international joint conference on biometrics, IJCB 2017, 2018 January, pp 22–30. https://doi.org/10.1109/BTAS.2017.8272678

Das S, Muhammad K, Bakshi S, Mukherjee I, Sa PK, Sangaiah AK, Bruno A (2019) Lip biometric template security framework using spatial steganography. Pattern Recognit Lett 126:102–110. https://doi.org/10.1016/j.patrec.2018.06.026

Deljavan Amiri M, Akhlaqian Tab F, Barkhoda W (2009) Retina identification based on the pattern of blood vessels using angular and radial partitioning. In: Lecture notes in computer science (including subseries lecture notes in artificial intelligence and lecture notes in bioinformatics), 5807 LNCS, pp 732–739. https://doi.org/10.1007/978-3-642-04697-1_68

Deng J (2017) Marginal loss for deep face recognition. In: Proceedings of the IEEE conference on computer vision and pattern recognition workshops

Deng J (2019) ArcFace: additive angular margin loss for deep face recognition. In: Proceedings of the IEEE conference on computer vision and pattern recognition, no 1

Deng J, Lu X (2019) Lightweight face recognition challenge. In: Proceedings of the international conference on computer vision workshop. https://doi.org/10.1109/ICCVW.2019.00322

Ding H, Zhou SK, Chellappa R (2017) FaceNet2ExpNet: regularizing a deep face recognition net for expression recognition. In: Proceedings of the12th IEEE international conference on automatic face & gesture recognition (FG 2017), pp 118–126. https://doi.org/10.1109/FG.2017.23

Dongyang DU, Lijun LU, Ruiyang FU, Lisha Y, Wufan C, Yaqin LIU (2019) Palm vein recognition based on end-to-end convolutional neural network 39(2):207–214. https://doi.org/10.12122/j.issn.1673-4254.2019.02.13

Donida Labati R, Genovese A, Muñoz E, Piuri V, Scotti F (2018) A novel pore extraction method for heterogeneous fingerprint images using convolutional neural networks. Pattern Recognit Lett 113:58–66. https://doi.org/10.1016/j.patrec.2017.04.001

Dua M, Gupta R, Khari M, Crespo RG (2019) Biometric iris recognition using radial basis function neural network. Soft Comput 23(22):11801–11815. https://doi.org/10.1007/s00500-018-03731-4

Duan Y, Lu J, Member S, Feng J, Zhou J (2017) Context-aware local binary feature learning for face recognition 8828:1–14. https://doi.org/10.1109/TPAMI.2017.2710183

Engineering C, Gables C (2017) Low resolution face recognition in surveillance systems using discriminant correlation analysis, pp 912–917. https://doi.org/10.1109/FG.2017.130

Fachrurrozi M (2017) Multi-object face recognition using content based image retrieval (CBIR), no x, pp 193–197

Fairuz MS, Habaebi MH, Elsheikh EMA (2019) Pre-trained based CNN model to identify finger vein. Bull Electr Eng Inform 8(3):855–862. https://doi.org/10.11591/eei.v8i3.1505

Fronitasari D, Indonesia U, Gunawan D, Indonesia U (2017) Palm vein recognition by using modified of local binary pattern (LBP) for extraction feature, pp 18–22

Gao Y, Zhao M, Yuille AL (2017). Semi-supervised sparse representation based classification for face recognition with insufficient labeled samples. https://doi.org/10.1109/TIP.2017.2675341

Gomez-barrero M, Kolberg J, Busch C (2018) Towards multi-modal finger presentation attack detection. In: 2018 14th international conference on signal-image technology & internet-based systems (SITIS), pp 547–552. https://doi.org/10.1109/SITIS.2018.00089

Gumede A, Viriri S, Gwetu M (2017) Hybrid component-based face recognition. In: Proceedings of the conference on information communication technology and society, pp 0–5

Guo X, Zhu E, Yin J (2018) A fast and accurate method for detecting fingerprint reference point. Neural Comput Appl 29(1):21–31. https://doi.org/10.1007/s00521-016-2285-9

Hamdan B, Mokhtar K (2016) Face recognition using angular radial transform. J King Saud Univ Comput Inf Sci. https://doi.org/10.1016/j.jksuci.2016.10.006

Harish M, Karthick R, Rajan RM, Vetriselvi V (2019) Iccce 2018. In: Proceedings of the international conference on communications and cyber physical engineering 2018, vol 500. https://doi.org/10.1007/978-981-13-0212-1

Hassner T, Sahin G, Medioni G, Masi I, Tu A (2019) Face-specific data augmentation for unconstrained face recognition. Int J Comput Vis 127:642–667. https://doi.org/10.1007/s11263-019-01178-0

Hatanaka Y, Tajima M, Kawasaki R, Saito K, Ogohara K, Muramatsu C, Fujita H (2017) Retinal biometrics based on iterative closest point algorithm. In: Proceedings of the annual international conference of the IEEE engineering in medicine and biology society, EMBS. https://doi.org/10.1109/EMBC.2017.8036840

Haware S, Barhatte A (2017) Retina based biometric identification using SURF and ORB feature descriptors. In: 2017 international conference on microelectronic devices, circuits and systems, ICMDCS 2017, 2017 January, pp 1–6. https://doi.org/10.1109/ICMDCS.2017.8211697

Heinsohn D, Villalobos E, Prieto L, Mery D (2019) Face recognition in low-quality images using adaptive sparse representations. Image Vis Comput 85:46–58. https://doi.org/10.1016/j.imavis.2019.02.012

Hemanth J (2018) Palm vein recognition based on competitive code, LBP and DCA fusion strategy

Hofbauer H, Jalilian E, Uhl A (2019) Exploiting superior CNN-based iris segmentation for better recognition accuracy. Pattern Recognit Lett 120:17–23. https://doi.org/10.1016/j.patrec.2018.12.021

Hosny KM, Elaziz MA (n.d.) Face recognition using exact Gaussian-hermit moments. https://doi.org/10.1007/978-3-030-03000-1

Hou B, Yan R (2019) Convolutional auto-encoder model for finger-vein verification. IEEE Trans Instrum Meas. https://doi.org/10.1109/tim.2019.2921135

Hsieh SH, Li YH, Wang W, Tien CH (2018) A novel anti-spoofing solution for iris recognition toward cosmetic contact lens attack using spectral ICA analysis. Sensors (Switzerland). https://doi.org/10.3390/s18030795

Hu Y, Sirlantzis K, Howells G (2017) A novel iris weight map method for less constrained iris recognition based on bit stability and discriminability. Image Vis Comput 58:168–180. https://doi.org/10.1016/j.imavis.2016.05.003

Huang J, Zhang Y, Zhang H, Cheng K (2019) Sparse representation face recognition based on gabor and CSLDP feature fusion. In: 2019 Chinese control and decision conference (CCDC), no 1, pp 5697–5701

Hussein ME, Spinoulas L, Xiong F, Abd-Almageed W (2019) Fingerprint presentation attack detection using a novel multi-spectral capture device and patch-based convolutional neural networks. In: 10th IEEE international workshop on information forensics and security. WIFS 2018:1–8. https://doi.org/10.1109/WIFS.2018.8630773

Ito K, Aoki T (2018) Recent advances in biometric recognition. ITE Trans Media Technol Appl 6(1):64–80. https://doi.org/10.3169/mta.6.64

Jacob IJ (2019) Capsule network based biometric recognition system 01(02):83–93

Jain A, East L, Ruud B, Pankanti S, Yorktown H (2002) Introduction to biometrics. Anil Jain Michigan State University. Biometrics: Personal Identification in Networked Society (ii), p 19

Jain AK, Kumar A (2012) Biometric recognition: an overview, pp 49–79. https://doi.org/10.1007/978-94-007-3892-8_3

Jalilian E, Uhl A (2019) Enhanced segmentation-CNN based finger-vein recognition by joint training with automatically generated and manual labels. In: 2019 IEEE 5th international conference on identity, security, and behavior analysis (ISBA), pp 1–8. https://doi.org/10.1109/isba.2019.8778522

Jang HU, Kim D, Mun SM, Choi S, Lee HK (2017) DeepPore: fingerprint pore extraction using deep convolutional neural networks. IEEE Signal Process Lett 24(12):1808–1812. https://doi.org/10.1109/LSP.2017.2761454

Jasim YA, Al-Ani AA, Al-Ani LA (2019) Iris recognition using principal component analysis. In: Proceedings—2018 1st annual international conference on information and sciences, AiCIS 2018, pp 89–95. https://doi.org/10.1109/AiCIS.2018.00028

Jeon WS, Rhee SY (2017) Fingerprint pattern classification using convolution neural network. Int J Fuzzy Logic Intell Syst 17(3):170–176. https://doi.org/10.5391/IJFIS.2017.17.3.170

Ji H, Sun Q, Ji Z, Yuan Y, Zhang G (2017) Collaborative probabilistic labels for face recognition from single sample per person. Pattern Recognit 62:125–134. https://doi.org/10.1016/j.patcog.2016.08.007

Joshi I, Anand A, Vatsa M, Singh R, Roy SD, Kalra PK (2019) Latent fingerprint enhancement using generative adversarial networks. In: Proceedings—2019 IEEE winter conference on applications of computer vision, WACV 2019, pp 895–903. https://doi.org/10.1109/WACV.2019.00100

Kang W, Lu Y, Li D, Jia W (2019) From noise to feature: Exploiting intensity distribution as a novel soft biometric trait for finger vein recognition. IEEE Trans Inf Forens Secur 14(4):858–869

Kaur B, Singh S, Kumar J (2018) Robust iris recognition using moment invariants. Wireless Pers Commun 99(2):799–828. https://doi.org/10.1007/s11277-017-5153-8

Keilbach P, Kolberg J, Gomez-Barrero M, Busch C, Langweg H (2018) Fingerprint presentation attack detection using laser speckle contrast imaging. In: 2018 international conference of the biometrics special interest group, BIOSIG 2018, pp 1–6. https://doi.org/10.23919/BIOSIG.2018.8552931

Kerrigan D, Trokielewicz M, Czajka A, Bowyer K (2019) Iris recognition with image segmentation employing retrained off-the-shelf deep neural networks. http://arxiv.org/abs/1901.01028

Khan T, Donald B, Khan M, Kong Y (1983) Efficient hardware implementation for fingerprint image enhancement using anisotropic gaussian filter. IEEE Trans Image Process 23(1):309–317. https://doi.org/10.1002/qua.560230126

Khan AI, Wani MA (2019) Patch-based segmentation of latent fingerprint images using convolutional neural network. Appl Artif Intell 33(1):87–100. https://doi.org/10.1080/08839514.2018.1526704

Khodadoust J, Khodadoust AM (2017) Fingerprint indexing based on minutiae pairs and convex core point. Pattern Recognit 67:110–126

Kim H, Cui X, Kim MG, Nguyen THB (2019) Fingerprint generation and presentation attack detection using deep neural networks. In: Proceedings—2nd international conference on multimedia information processing and retrieval, MIPR 2019, pp 375–378. https://doi.org/10.1109/MIPR.2019.00074

Kr P, Lenc L (2019) Novel texture descriptor family for face recognition 2:37–47. https://doi.org/10.1007/978-3-030-20915-5

Kr P, Lenc L (2019) Enhanced local binary patterns for automatic face recognition 2:27–36. https://doi.org/10.1007/978-3-030-20915-5

Král P, Lenc L and Vrba A (2019) Enhanced local binary patterns for automatic face recognition. In: International conference on artificial intelligence and soft computing. Springer, Cham, pp 27–36

Krish RP, Fierrez J, Ramos D, Alonso-Fernandez F, Bigun J (2019) Improving automated latent fingerprint identification using extended minutia types. Inf Fusion 50:9–19. https://doi.org/10.1016/j.inffus.2018.10.001

Kumar SVM, Nishanth R, Sani N, Joseph AJ, Martin A (2019) Specular reflection removal using morphological filtering for accurate iris recognition. In: 6th IEEE international conference on smart structures and systems, ICSSS 2019, pp 1–4. https://doi.org/10.1109/ICSSS.2019.8882863

Lee EC, Jung H, Kim D (2011) New finger biometric method using near infrared imaging. Sensors 11(3):2319–2333. https://doi.org/10.3390/s110302319

Li C (2017) Dependence structure of gabor wavelets for face recognition, pp 0–4

Li J, Feng J, Kuo CCJ (2018) Deep convolutional neural network for latent fingerprint enhancement. Signal Process Image Commun 60:52–63. https://doi.org/10.1016/j.image.2017.08.010

Li L, Ge H, Tong Y, Zhang Y (2017). Face recognition using gabor-based feature extraction and feature space transformation fusion method for single image per person problem. https://doi.org/10.1007/s11063-017-9693-4

Li G, Kim J (2016) Author’s accepted manuscript palmprint recognition with local micro-structure tetra pattern reference. Pattern Recognit. https://doi.org/10.1016/j.patcog.2016.06.025

Lin BS, Yao YH, Liu CF, Lien CF, Lin BS (2017) Impact of the lips for biometrics. IEEE Access 5:794–801. https://doi.org/10.1109/ACCESS.2017.2649838

Linsangan NB, Panganiban AG, Flores PR, Poligratis HAT, Victa AS, Torres JL, Villaverde J (2019) Real-time iris recognition system for non-ideal iris images. ACM Int Conf Proc Ser. https://doi.org/10.1145/3313991.3314002

Liu JLC, Chen X, Zhou J, Tan T, Zheng N, Zha H, Hutchison D (2018) Pattern recognition and computer vision

Liu X, Kan M, Wu W, Shan S (2017) VIPLFaceNet : an open source deep face recognition SDK 11(2):208–218

Liu N, Liu J, Sun Z, Tan T (2017) A code-level approach to heterogeneous iris recognition. IEEE Trans Inf Forens Secur 12(10):2373–2386. https://doi.org/10.1109/TIFS.2017.2686013

Liu Y, Member S, Lin C, Guo J, Member S (2012) Impact of the lips in facial biometrics 21(6):3092–3101

Liu H, Yang L, Yang G, Yin Y (2018) Discriminative binary descriptor for finger vein recognition. IEEE Access 6:5795–5804. https://doi.org/10.1109/ACCESS.2017.2787543

Liu H, Yang G, Yang L, Yin Y (2019) Learning personalized binary codes for finger vein recognition. Neurocomputing 365:62–70. https://doi.org/10.1016/j.neucom.2019.07.057

Liu M, Zhou Z, Shang P, Xu D (2019) Fuzzified image enhancement for deep learning in iris recognition. IEEE Trans Fuzzy Syst. https://doi.org/10.1109/tfuzz.2019.2912576

Llano EG, García Vázquez MS, Vargas JMC, Fuentes LMZ, Ramírez Acosta AA (2018) Optimized robust multi-sensor scheme for simultaneous video and image iris recognition. Pattern Recognit Lett 101:44–51. https://doi.org/10.1016/j.patrec.2017.11.012

Lu J, Member S, Liong VE, Member S, Zhou J (2017) Simultaneous local binary feature learning and encoding for homogeneous and heterogeneous face recognition 8828:1–14. https://doi.org/10.1109/TPAMI.2017.2737538

Lu Yu, Xie S, Wu S (2019) Exploring competitive features using deep convolutional neural network for finger vein recognition. IEEE Access 7:35113–35123. https://doi.org/10.1109/ACCESS.2019.2902429

Lu Y, Yan J, Gu K (2018) Review on automatic lip reading techniques. Int J Pattern Recognit Artif Intell 32(7):1–21. https://doi.org/10.1142/S0218001418560074

Lv J (2018) A new discriminative collaborative neighbor representation method for robust face recognition, pp 74713–74727

Lv J-J, Shao X, Huang J, Zhou X, Zhou X, Lv J, Zhou X (2016) Face-specific data augmentation for unconstrained face recognition. Neurocomputing. https://doi.org/10.1016/j.neucom.2016.12.025

Malik F, Azis A, Nasrun M, Setianingsih C, Murti MA (2018) Face recognition in night day using method eigenface, pp 103–108

Manickam A, Devarasan E, Manogaran G, Priyan MK, Varatharajan R, Hsu CH, Krishnamoorthi R (2019) Score level based latent fingerprint enhancement and matching using SIFT feature. Multimed Tools Appl 78(3):3065–3085. https://doi.org/10.1007/s11042-018-5633-1

Mariño C, Penedo MG, Penas M, Carreira MJ, Gonzalez F (2006) Personal authentication using digital retinal images. Pattern Anal Appl 9(1):21–33. https://doi.org/10.1007/s10044-005-0022-6

Marra F, Poggi G, Sansone C, Verdoliva L (2018) A deep learning approach for iris sensor model identification. Pattern Recognit Lett 113:46–53. https://doi.org/10.1016/j.patrec.2017.04.010

Mathematics A (2018) Biometric retinal security system for user identification and authentication in smartphones 119(14):187–202

Mazumdar JB (2018) Retina based biometric authentication system: a review. Int J Adv Res Comput Sci 9(1):711–718. https://doi.org/10.26483/ijarcs.v9i1.5322

Mazumdar JB, Nirmala SR (2018) Retina based biometric authentication system: a review. Int J Adv Res Comp Sci 9(1):711–718

Mazumdar J, Nirmala SR (2019) Person identification using parabolic model-based algorithm in color retinal images. https://doi.org/10.18178/ijeetc.8.6.358-366

Meng XJ, Yin YL, Yang GP, Xi XM (2013) Retinal identification based on an improved circular gabor filter and scale invariant feature transform. Sensors (Switzerland) 13(7):9248–9266. https://doi.org/10.3390/s130709248

Minaee S, Abdolrashidi A (2019) DeepIris: iris recognition using a deep learning approach. http://arxiv.org/abs/1907.09380

Mir SA, Khan S, Bhat MA, Mehraj H (2018) Person identification by lips using SGLDM and support vector machine 4(1):152–157

Miura N, Nagasaka A, Miyatake T (2004) Feature extraction of finger vein patterns based on iterative line tracking and its application to personal identification. Syst Comput Jpn 35(7):61–71. https://doi.org/10.1002/scj.10596

Miura N, Nagasaka A, Miyatake T (2007) Extraction of finger-vein patterns using maximum curvature points in image profiles. IEICE Trans Inf Syst E90-D(8):1185–1194. https://doi.org/10.1093/ietisy/e90-d.8.1185

Moorfield J, Wang S, Yang W, Bedari A, Van Der Kamp P (2020) A Möbius transformation based model for fingerprint minutiae variations. Pattern Recognit 98:107054. https://doi.org/10.1016/j.patcog.2019.107054

Morales S, Naranjo V, Angulo J, Legaz-Aparicio AG, Verdú-Monedero R (2017) Retinal network characterization through fundus image processing: significant point identification on vessel centerline. Signal Process Image Commun 59:50–64. https://doi.org/10.1016/j.image.2017.03.013

Mura V, Orru G, Casula R, Sibiriu A, Loi G, Tuveri P et al. (2018) LivDet 2017 fingerprint liveness detection competition 2017. In: Proceedings—2018 international conference on biometrics, ICB 2018, pp 297–302. https://doi.org/10.1109/ICB2018.2018.00052

Nachar R, Inaty E, Bonnin PJ, Alayli Y (2020) Hybrid minutiae and edge corners feature points for increased fingerprint recognition performance. Pattern Anal Appl 23(1):213–222

Naseem I, Aleem A, Togneri R, Bennamoun M (2017) Iris recognition using class-specific dictionaries. Comput Electr Eng 62:178–193. https://doi.org/10.1016/j.compeleceng.2015.12.017

Nguyen DL, Cao K, Jain AK (2019) Automatic latent fingerprint segmentation. In: 2018 IEEE 9th international conference on biometrics theory, applications and systems, BTAS 2018, pp 1–9. https://doi.org/10.1109/BTAS.2018.8698544

Nguyen DL, Cao K, Jain AK (2018) Robust minutiae extractor: integrating deep networks and fingerprint domain knowledge. In: Proceedings—2018 international conference on biometrics, ICB 2018, pp 9–16. https://doi.org/10.1109/ICB2018.2018.00013

Nguyen K, Fookes C, Jillela R, Sridharan S, Ross A (2017) Long range iris recognition: a survey. Pattern Recognit 72:123–143. https://doi.org/10.1016/j.patcog.2017.05.021

Nguyen K, Fookes C, Ross A, Sridharan S (2017) Iris recognition with off-the-shelf CNN features: a deep learning perspective. IEEE Access 6:18848–18855. https://doi.org/10.1109/ACCESS.2017.2784352

Oliveira A, Pereira S, Silva CA (2018) Retinal vessel segmentation based on fully convolutional neural networks. Expert Syst Appl 112:229–242. https://doi.org/10.1016/j.eswa.2018.06.034

Pandya B, Cosma G, Alani AA, Taherkhani A, Bharadi V, McGinnity TM (2018) Fingerprint classification using a deep convolutional neural network. In: 2018 4th international conference on information management, ICIM 2018, pp 86–91. https://doi.org/10.1109/INFOMAN.2018.8392815

Panetta K, Kamath SKM, Rajeev S, Agaian SS (2019) LQM: localized quality measure for fingerprint image enhancement. IEEE Access 7:104567–104576. https://doi.org/10.1109/access.2019.2931980

Parihar RS, Jain S (2019) A robust method to recognize palm vein using SIFT and SVM classifier 1:1703–1710

Park Y, Jang U, Lee EC (2018) Statistical anti-spoofing method for fingerprint recognition. Soft Comput 22(13):4175–4184. https://doi.org/10.1007/s00500-017-2707-3

Park K, Song M, Youn Kim S (2018) The design of a single-bit CMOS image sensor for iris recognition applications. Sensors (Switzerland). https://doi.org/10.3390/s18020669

Pei T, Zhang L, Wang B (2016) Decision Pyramid Classifier for face recognition under complex variations using single sample per person. Pattern Recognit. https://doi.org/10.1016/j.patcog.2016.11.016

Peng C, Gao X, Wang N, Li J (2018) Face recognition from multiple stylistic sketches: scenarios, datasets, and evaluation. Pattern Recognit 84:262–272. https://doi.org/10.1016/j.patcog.2018.07.014

Peralta D, García S, Benitez JM, Herrera F (2017) Minutiae-based fingerprint matching decomposition: methodology for big data frameworks. Inf Sci 408:198–212. https://doi.org/10.1016/j.ins.2017.05.001

Piciucco E, Maiorana E, Campisi P (2017) Biometric fusion for palm-vein-based recognition systems palm vein biometric recognition: state of the art, pp 18–28. https://doi.org/10.1007/978-3-319-67639-5

Piciucco E, Maiorana E, Campisi P (2018) Palm vein recognition using a high dynamic range approach, pp 1–8. https://doi.org/10.1049/iet-bmt.2017.0192

Qin H, He X, Yao X, Li H (2017) Finger-vein verification based on the curvature in Radon space. Expert Syst Appl 82:151–161. https://doi.org/10.1016/j.eswa.2017.03.068

Qin H, Wang P (2019) Finger-vein verification based on LSTM recurrent neural networks. Appl Sci (Switzerland) 9(8):1–18. https://doi.org/10.3390/app9081687

Qureshi I, Khan MA, Sharif M, Saba T, Ma J (2020) Detection of glaucoma based on cup-to-disc ratio using fundus images 19(1):1–16

Qureshi I, Ma J, Abbas Q (2019) Recent development on detection methods for the diagnosis of diabetic retinopathy. Symmetry 11(6):1–34. https://doi.org/10.3390/sym11060749

Qureshi I, Ma J, Shaheed K (2019) A hybrid proposed fundus image enhancement framework for diabetic retinopathy, pp 1–16. https://doi.org/10.3390/a12010014

Raja KB, Raghavendra R, Venkatesh S, Busch C (2017) Multi-patch deep sparse histograms for iris recognition in visible spectrum using collaborative subspace for robust verification. Pattern Recognit Lett 91:27–36. https://doi.org/10.1016/j.patrec.2016.12.025

Ramachandra R, Raja KB, Venkatesh SK, Busch C (2019) Design and development of low-cost sensor to capture ventral and dorsal finger vein for biometric authentication. IEEE Sens J 19(15):6102–6111. https://doi.org/10.1109/JSEN.2019.2906691

Rana HK (2017) SM Gr up SM journal of iris recognition system using PCA based on DWT. SM J Biometrics Biostat 2:1015. https://doi.org/10.5281/zenodo.2580202

Ranjani R, Priya C (2018) A survey on face recognition techniques: a review. Int J Pure Appl Math 118:253–274

Rathgeb C, Wagner J, Busch C (2019) SIFT-based iris recognition revisited: prerequisites, advantages and improvements. Pattern Anal Appl 22(3):889–906. https://doi.org/10.1007/s10044-018-0719-y

Roy ND, Biswas A (2019) Fast and robust retinal biometric key generation using deep neural nets

Rui Z, Yan Z (2019) A survey on biometric authentication: toward secure and privacy-preserving identification. IEEE Access 7:5994–6009. https://doi.org/10.1109/ACCESS.2018.2889996

Sahu B, Kumar Sa P, Bakshi S, Sangaiah AK (2018) Reducing dense local feature key-points for faster iris recognition. Comput Electr Eng 70:939–949. https://doi.org/10.1016/j.compeleceng.2017.12.048

Sang J, Wang H, Qian Q, Wu H, Chen Y (2018) An efficient fingerprint identification algorithm based on minutiae and invariant moment. Pers Ubiquit Comput 22(1):71–80. https://doi.org/10.1007/s00779-017-1094-1

De Santis M, Agnelli S, Don V, Gnocchi C (2017) 3D ultrasound palm vein recognition through the centroid method for biometric purposes, pp 1–4

Shaheed K, Liu H, Yang G, Qureshi I, Gou J, Yin Y (2018) A systematic review of finger vein recognition techniques. Information (Switzerland). https://doi.org/10.3390/info9090213

Shaheed K, Yang L, Yang G, Qureshi I, Yin Y (2018) Novel image quality assessment and enhancement techniques for finger vein recognition, pp 223–231

Shao L, Zhu R, B QZ (2016) A finger vein identification system based on image quality assessment, vol 3, pp 711–719. https://doi.org/10.1007/978-3-319-46654-5

Sharma RP, Dey S (2019). Quality analysis of fingerprint images using local phase quantization. https://doi.org/10.1007/978-3-030-29888-3_53

Shuyi L, Haigang Z, Jinfeng Y (2019) Finger vein recognition based on local graph structural coding and CNN. May, vol 8. https://doi.org/10.1117/12.2524152

Soh SC, Ibrahim MZ, Abas MF (2019) Image fusion based multi resolution and frequency partition discrete cosine transform for palm vein recognition. In: 2019 IEEE 6th international conference on industrial engineering and applications (ICIEA), pp 367–371

Soh SC, Ibrahim MZ, Yakno MB, Mulvaney DJ (2017) Palm vein recognition using scale invariant feature transform with RANSAC mismatching removal. IT Converg Secur 2017:2. https://doi.org/10.1007/978-981-10-6451-7

Song JM, Kim W, Park KR (2019) Finger-vein recognition based on deep densenet using composite image. IEEE Access 7:66845–66863. https://doi.org/10.1109/ACCESS.2019.2918503

de Souza GB, Santos DFdS, Pires RG, Marana AN, Papa JP (2018) Deep features extraction for robust fingerprint spoofing attack detection. J Artif Intell Soft Comput Res 9(1):41–49. https://doi.org/10.2478/jaiscr-2018-0023

Sun Y, Wang X, Tang X (2013) Deep convolutional network cascade for facial point detection. https://doi.org/10.1109/CVPR.2013.446

Sundararajan K, Woodard DL (2018) Deep learning for biometrics: a survey. ACM Comput Surv. https://doi.org/10.1145/3190618

Sutanty E, Rahayu DA, Rodiah, Susetianingtias DT, Madenda S (2017) Retinal blood vessel segmentation and bifurcation detection using combined filters. In: Proceeding—2017 3rd international conference on science in information technology: theory and application of IT for education, industry and society in big data era, ICSITech 2017, 2018 January, pp 563–567. https://doi.org/10.1109/ICSITech.2017.8257176

Svoboda J, Monti F, Bronstein MM (2018) Generative convolutional networks for latent fingerprint reconstruction. In: IEEE international joint conference on biometrics, IJCB 2017, 2018 January, pp 429–436. https://doi.org/10.1109/BTAS.2017.8272727

Syarif MA, Ong TS, Teoh ABJ, Tee C (2017) Enhanced maximum curvature descriptors for finger vein verification. Multimed Tools Appl 76(5):6859–6887. https://doi.org/10.1007/s11042-016-3315-4

Tams B (2013) Absolute fingerprint pre-alignment in minutiae-based cryptosystems. In: BIOSIG 2013—proceedings of the 12th international conference of the biometrics special interest group, pp 1–12

Tang Z, Wu X, Fu B, Chen W, Feng H (2018) Fast face recognition based on fractal theory. Appl Math Comput 321:721–730. https://doi.org/10.1016/j.amc.2017.11.017

Article   MathSciNet   MATH   Google Scholar  

Tann H, Zhao H, Reda S (2019) A resource-efficient embedded iris recognition system using fully convolutional networks. ACM J Emerg Technol Comput Syst 16(1):1–23. https://doi.org/10.1145/3357796

Tertychnyi P, Ozcinar C, Anbarjafari G (2018) Low-quality fingerprint classification using deep neural network. IET Biometrics 7(6):550–556. https://doi.org/10.1049/iet-bmt.2018.5074

Thapar D, Jaswal G, Nigam A (2018) PVSNet: palm vein authentication siamese network trained using triplet loss and adaptive hard mining by learning enforced domain specific features. In: 2019 IEEE 5th international conference on identity, security, and behavior analysis (ISBA), pp 1–8

Thompson J, Flynn P, Boehnen C, Santos-Villalobos H (2019) Assessing the impact of corneal refraction and iris tissue non-planarity on iris recognition. IEEE Trans Inf Forens Secur 14(8):2102–2112. https://doi.org/10.1109/TIFS.2018.2869342

Tran MH, Duong TN, Nguyen DM, Dang QH (2017) A local feature vector for an adaptive hybrid fingerprint matcher. In: Proceedings of KICS-IEEE international conference on information and communications with samsung LTE and 5G special workshop, ICIC 2017, pp 249–253. https://doi.org/10.1109/INFOC.2017.8001668

Travieso CM, Ravelo-García AG, Alonso JB, Canino-Rodríguez JM, Dutta MK (2019) Improving the performance of the lip identification through the use of shape correction. Appl Intell 49(5):1823–1840

Travieso CM, Zhang J, Miller P, Alonso JB (2014) Using a discrete Hidden Markov Model Kernel for lip-based biometric identification. Image Vis Comput 32(12):1080–1089. https://doi.org/10.1016/j.imavis.2014.10.001

Unar JA, Seng WC, Abbasi A (2014) A review of biometric technology along with trends and prospects. Pattern Recognit 47(8):2673–2688. https://doi.org/10.1016/j.patcog.2014.01.016

Wan C, Wang L, Phoha VV (2018) A survey on gait recognition. ACM Comput Surv. https://doi.org/10.1145/3230633

Wang X (2019) Palm vein recognition based on competitive code and, pp 179–183

Wang K, Kumar A (2019) Toward more accurate iris recognition using dilated residual features. IEEE Trans Inf Forens Secur 14(12):3233–3245

Wang Z, Ma S, Han M, Hu G (2017) Long-distance/environment face image enhancement method for recognition, no 1, pp 501–511. https://doi.org/10.1007/978-3-319-71607-7

Wang H, Yang X, Ma L, Liang R (2017) Fingerprint pore extraction using U-Net based fully convolutional network 1:474–483. https://doi.org/10.1007/978-3-319-69923-3

Xi X, Yang L (2017) Learning discriminative binary codes for finger vein recognition. Pattern Recognit 66:26–33. https://doi.org/10.1016/j.patcog.2016.11.002

Xia Z, Lv R, Zhu Y, Ji P, Sun H, Shi YQ (2017) Fingerprint liveness detection using gradient-based texture features. SIViP 11(2):381–388. https://doi.org/10.1007/s11760-016-0936-z

Xin M (2017) Palm vein recognition method based on fusion of local Gabor histograms. J China Univ Posts Telecommun 24(6):55–66. https://doi.org/10.1016/S1005-8885(17)60242-5

Yahaya YH, Shamsuddin SM, Leng WY, Technology D, Pertahanan U (2016) Finger vein feature extraction using discretization. November, pp 28–29

Yang W, Huang X, Zhou F, Liao Q (2014) Comparative competitive coding for personal identification by using finger vein and finger dorsal texture fusion. Inf Sci 268:20–32. https://doi.org/10.1016/j.ins.2013.10.010

Yang W, Ji W, Xue JH, Ren Y, Liao Q (2019) A hybrid finger identification pattern using polarized depth-weighted binary direction coding. Neurocomputing 325:260–268. https://doi.org/10.1016/j.neucom.2018.10.042

Yang W, Ma G, Li W, Liao Q (2013) Finger vein verification based on neighbor pattern coding. IEICE Trans Inf Syst E96-D(5):1227–1229. https://doi.org/10.1587/transinf.E96.D.1227

Yang J, Shi Y, Jia G (2017) Finger-vein image matching based on adaptive curve transformation. Pattern Recognit 66:34–43. https://doi.org/10.1016/j.patcog.2017.01.008

Yang W, Wang S, Hu J, Zheng G, Valli C (2019) Security and accuracy of fingerprint-based biometrics: a review. Symmetry. https://doi.org/10.3390/sym11020141

Yang W, Wang S, Hu J, Zheng G, Yang J, Valli C (2019) Securing deep learning based edge finger vein biometrics with binary decision diagram. IEEE Trans Ind Inf 15(7):4244–4253. https://doi.org/10.1109/TII.2019.2900665

Yang M, Wang X, Zeng G, Shen L (2017) Joint and collaborative representation with local adaptive convolution feature for face recognition with single sample per person. Pattern Recognit 66:117–128. https://doi.org/10.1016/j.patcog.2016.12.028

Yang J, Wei J, Shi Y (2019) Accurate ROI localization and hierarchical hyper-sphere model for finger-vein recognition. Neurocomputing 328:171–181. https://doi.org/10.1016/j.neucom.2018.02.098

Yang L, Yang G, Wang K, Liu H, Xi X, Yin Y (2019) Point grouping method for finger vein recognition. IEEE Access 7:28185–28195. https://doi.org/10.1109/ACCESS.2019.2901017

Yang L, Yang G, Xi X, Meng X, Zhang C, Yin Y (2017) Tri-branch vein structure assisted finger vein recognition. IEEE Access 5:21020–21028. https://doi.org/10.1109/ACCESS.2017.2728797

Yang L, Yang G, Yin Y, Xi X (2018) Finger vein recognition with anatomy structure analysis. IEEE Trans Circuits Syst Video Technol 28(8):1892–1905. https://doi.org/10.1109/TCSVT.2017.2684833

Ye L, Wang H, Du M, He Y, Tao L (2019) Weber local descriptor with edge detection and double Gabor orientations for finger vein recognition. In: Tenth international conference on graphics and image processing (ICGIP 2018), vol 11069. International Society for Optics and Photonics, p 110693J

Yin B, Tran L, Li H, Shen X, Liu X (2018) Towards interpretable face recognition, pp 9348–9357

You W, Zhou W, Huang J, Yang F, Liu Y, Chen Z (2019) A bilayer image restoration for finger vein recognition. Neurocomputing 348:54–65. https://doi.org/10.1016/j.neucom.2018.06.085

Yuan X, Gu L, Chen T, Elhoseny M, Wang W (2018) A fast and accurate retina image verification method based on structure similarity. In: Proceedings—IEEE 4th international conference on big data computing service and applications, BigDataService 2018, pp 181–185. https://doi.org/10.1109/BigDataService.2018.00034

Yuan C, Li X, Wu QMJ, Li J, Sun X (2017) Fingerprint liveness detection from different fingerprint materials using convolutional neural network and principal component analysis. Comput Mater Continua 53(4):357–371

Yuan C, Sun X, Wu QMJ (2019) Difference co-occurrence matrix using BP neural network for fingerprint liveness detection. Soft Comput 23(13):5157–5169. https://doi.org/10.1007/s00500-018-3182-1

Zafar U, Ghafoor M, Zia T, Ahmed G, Latif A, Malik KR, Sharif AM (2019) Face recognition with Bayesian convolutional networks for robust surveillance systems

Zhang M, He Z, Zhang H, Tan T, Sun Z (2019) Toward practical remote iris recognition: a boosting based framework. Neurocomputing 330:238–252. https://doi.org/10.1016/j.neucom.2017.12.053

Zhang Y, Li W, Zhang L, Lu Y (2019) Adaptive gabor convolutional neural networks for finger-vein recognition. In: 2019 International Conference on High Performance Big Data and Intelligent systems, HPBD and IS 2019, (61572458), pp 219–222. https://doi.org/10.1109/HPBDIS.2019.8735471

Zhang Y, Shi D, Zhan X, Cao D, Zhu K, Li Z (2019) Slim-ResCNN: a deep residual convolutional neural network for fingerprint liveness detection. IEEE Access 7:91476–91487. https://doi.org/10.1109/access.2019.2927357

Zhao Z, Kumar A (2017) Towards more accurate iris recognition using deeply learned spatially corresponding features. In: Proceedings of the IEEE international conference on computer vision, 2017 October, pp 3829–3838. https://doi.org/10.1109/ICCV.2017.411

Zhao T, Liu Y, Huo G, Zhu X (2019) A deep learning iris recognition method based on capsule network architecture. IEEE Access 7:49691–49701. https://doi.org/10.1109/ACCESS.2019.2911056

Download references

Acknowledgements

The work of this paper is financially supported by NSF of Guangdong Province (No. 2019A1515010833) and fundamental Research Funds for the Central Universities (No. 2020ZYGXZR089).

Author information

Authors and affiliations.

School of Computer Science and Engineering, South China University of Technology, Guangzhou, 510006, China

Kashif Shaheed, Aihua Mao & Xingming Zhang

Key Laboratory of Space Photoelectric Detection and Perception, Ministry of Industry and Information Technology and College of Astronautics, Nanjing University of Aeronautics and Astronautics, Nanjing, 211106, Jiangsu, China

Imran Qureshi

Department of Computational Sciences, Maharaja Ranjit Singh Punjab Technical University, Bathinda, Punjab, 151001, India

Munish Kumar

College of Computer and Information Sciences, Imam Muhammad Ibn Saud Islamic University (IMSIU), Riyadh, 11432, Saudi Arabia

Qaisar Abbas

School of Software Engineering, Shandong University, Jinan, 250100, China

You can also search for this author in PubMed   Google Scholar

Corresponding authors

Correspondence to Kashif Shaheed or Aihua Mao .

Ethics declarations

Conflict of interest.

The authors declare that they have no conflict of interest.

Additional information

Publisher's note.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Reprints and permissions

About this article

Shaheed, K., Mao, A., Qureshi, I. et al. A Systematic Review on Physiological-Based Biometric Recognition Systems: Current and Future Trends. Arch Computat Methods Eng 28 , 4917–4960 (2021). https://doi.org/10.1007/s11831-021-09560-3

Download citation

Received : 12 August 2020

Accepted : 28 January 2021

Published : 23 February 2021

Issue Date : December 2021

DOI : https://doi.org/10.1007/s11831-021-09560-3

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

• Find a journal
• Publish with us
• Track your research

• Frontiers in Artificial Intelligence
• Pattern Recognition
• Research Topics

Deep Learning for Biometrics: Advances and Applications

Total Downloads

Total Views and Downloads

About this Research Topic

Analyzing people’s unique physical and behavioral characteristics is the essence of the science of biometrics. Over the last two decades, we have witnessed an exponential growth of research interests in this domain, through which biometrics has surged from interesting and conventional pattern recognition applications, to deep learning-based mainstream research topics. Three major reasons account for this, first, the technological progress of the sensors that capture biometric signals, second, the significant increase in machines' computing power, and third, the democratization of deep learning paradigms. In this article collection, we are focusing on deep learning and biometric research topics that continue to be challenging, including evaluating new biometric techniques, remarkably improving the performance of existing ones, and ensuring the scalability of biometric systems to handle the ever-increasing amount of biometric data. In fact, biometric systems follow a typical pipeline that is composed of separate acquisition, preprocessing, feature extraction and classification. Deep learning as a data-driven representation learning approach has been shown to be a promising alternative to conventional data-agnostic and handcrafted pre-processing and feature extraction for biometric systems. Combining deep learning models with biological visual perception, many biologically plausible approaches to deep learning have been proposed for review. Including research on a single deep learning model including CNN, RNN, AE, GAN, GNN, and reinforcement learning models, the recent approaches of visual perception computational models oriented deep learning, showing its advantage and its progressive impact on artificial intelligence. Topics include but are not restricted to the following: • AI-Based Biometric Applications • Identity, Expression, Gender and Age Recognition • Vision and Perception • Deep Learning Techniques and Intelligent Systems for analyzing biometric data, such as CNN, RNN, transfer learning with convolutional neural networks, GAN, GNN, and reinforcement learning models for face recognition, gender and ethnicity classification, etc. • Behavioral Analysis and Information Inference • Activity, Action and Posture Recognition • Biometric systems for information assurance • Privacy, Security and Access Control

Keywords : biometrics, deep learning, generative adversarial networks, convolutional neural networks, recurrent neural networks, identity recognition, gender classification, privacy

Important Note : All contributions to this Research Topic must be within the scope of the section and journal to which they are submitted, as defined in their mission statements. Frontiers reserves the right to guide an out-of-scope manuscript to a more suitable section or journal at any stage of peer review.

Topic Editors

Topic coordinators, submission deadlines, participating journals.

Manuscripts can be submitted to this Research Topic via the following journals:

total views

• Demographics

No records found

total views article views downloads topic views

Top countries

Top referring sites, about frontiers research topics.

With their unique mixes of varied contributions from Original Research to Review Articles, Research Topics unify the most influential researchers, the latest key findings and historical advances in a hot research area! Find out more on how to host your own Frontiers Research Topic or contribute to one as an author.

An official website of the United States government

The .gov means it’s official. Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

The site is secure. The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

• Publications
• Account settings

Preview improvements coming to the PMC website in October 2024. Learn More or Try it out now .

• Advanced Search
• Journal List
• Sensors (Basel)

Biometric Recognition: A Systematic Review on Electrocardiogram Data Acquisition Methods

Teresa m. c. pereira.

1 IEETA, DETI, LASI, Universidade de Aveiro, 3810-193 Aveiro, Portugal

Raquel C. Conceição

2 Instituto de Biofísica e Engenharia Biomédica, Faculdade de Ciências da Universidade de Lisboa, Campo Grande, 1749-016 Lisboa, Portugal

Vitor Sencadas

3 Instituto de Materiais (CICECO), Departamento de Materiais e Cerâmica, Universidade de Aveiro, 3810-193 Aveiro, Portugal

Raquel Sebastião

Associated data.

Not applicable.

In the last decades, researchers have shown the potential of using Electrocardiogram (ECG) as a biometric trait due to its uniqueness and hidden nature. However, despite the great number of approaches found in the literature, no agreement exists on the most appropriate methodology. This paper presents a systematic review of data acquisition methods, aiming to understand the impact of some variables from the data acquisition protocol of an ECG signal in the biometric identification process. We searched for papers on the subject using Scopus, defining several keywords and restrictions, and found a total of 121 papers. Data acquisition hardware and methods vary widely throughout the literature. We reviewed the intrusiveness of acquisitions, the number of leads used, and the duration of acquisitions. Moreover, by analyzing the literature, we can conclude that the preferable solutions include: (1) the use of off-the-person acquisitions as they bring ECG biometrics closer to viable, unconstrained applications; (2) the use of a one-lead setup; and (3) short-term acquisitions as they required fewer numbers of contact points, making the data acquisition of benefit to user acceptance and allow faster acquisitions, resulting in a user-friendly biometric system. Thus, this paper reviews data acquisition methods, summarizes multiple perspectives, and highlights existing challenges and problems. In contrast, most reviews on ECG-based biometrics focus on feature extraction and classification methods.

1. Introduction

Nowadays, recognition systems are used in a variety of real-world applications to protect ourselves and our information. While some of these systems still depend on conventional technologies, such as cards, keys, or passwords, these mechanisms often have usability and security issues.

As a result, there has been a recent interest in the biometric field. Biometric recognition uses distinctive physiological and behavioral characteristics to automatically identify individuals. The former characteristics can include the face, fingerprint, iris, and hand geometry, whereas the latter can be gait signature and keystroke [ 1 ].

In recent years, researchers have been exploring the use of electrocardiogram (ECG) signals as a biometric recognition trait due to their unique properties: (1) liveness detection: Since the ECG is a recording of the electrical activity of the heart, ECG signals can only be acquired from living individuals [ 2 ]; (2) high security: ECG signals are extremely difficult to counterfeit and consequently, a technology to artificially produce them has not been developed yet [ 3 ]; (3) combined information: the analysis of ECG signals can give us information regarding the identity of a person, as well as heart conditions and emotional and physical status [ 4 ]. The most important advantage of ECG signals is their uniqueness among individuals, which is mainly due to changes in ionic potential, the levels of electrolytes in the plasma, and physiological differences caused by chest geometry, size, and position of the heart [ 2 ]. A typical ECG wave, such as the one presented in Figure 1 , consists of a P wave, a QRS complex, and a T wave.

The sequence of depolarization and repolarization events in the heart and their relationship with the different heartbeat waveforms in an ECG signal (adapted from [ 5 ], original figure kindly provided by the authors).

A biometric system is a technology that identifies or authenticates a person through their unique biometric traits. It consists of three main components: an acquisition module, which consists of a sensor that measures the biometric trait; a storage module, where personal data of enrolled subjects is stored); and a biometric algorithm. The biometric algorithm processes the data from the acquisition and storage modules, following two steps: feature extraction and pattern recognition [ 5 ]. Concerning a biometric system using ECG, signals can be acquired through different formats. The standard 12-lead ECG provides information on cardiac activity from 12 different leads over a short period of time, while Holter ECGs record electrical activity from five to seven leads over longer periods of time. Although 12-lead ECGs provide more information, they are not practical for real-world use. Instead, off-the-person methods that acquire ECG signals through skin or finger contact have become more common, making the process more convenient for users [ 6 ].

The advancement of sensing technology has made it possible to explore the use of ECG as a non-invasive biometric, similar to a fingerprint. This has made society’s acceptance of ECG as a biometric very promising [ 2 ]. In addition to traditional off-the-person methods, small wireless ECG body sensors are being developed for long-term monitoring. These sensors use a single lead to measure the electrical potential difference between electrodes placed near the heart. These sensors allow ECG analysis and monitoring to be used for a wider range of applications beyond diagnosing cardiovascular disorders [ 6 ]. However, when compared to medical devices such as Holter devices, wearable sensors produce noisier signals due to various factors, such as the type of electrodes and the number and location of leads. While medical ECG recorders use 12 or 6 wet electrodes, wearable devices typically use between one and three dry electrodes, with only the first lead being used due to its easy implementation in mobile devices. Medical ECG recorders generally provide more reliable data than wearable devices due to their longer and more detailed recording periods and the higher complexity of the setup [ 7 ].

In the next stage of the biometric process, features are extracted from the ECG data. These features are specific attributes of the ECG that allow for the recognition of a particular individual based on inter-subject variability. Feature extraction is a crucial step in pattern recognition. Approaches for feature extraction can be divided into three categories: fiducial, non-fiducial, and hybrid (or partially fiducial). Fiducial-based techniques rely on the accurate detection of reference points, such as the P wave, QRS complex, and T wave. These techniques can also use interval, amplitude, angle, and area measurements of these points as biometric features. However, these approaches require a lot of feature engineering, which can be time-consuming [ 8 , 9 ]. Non-fiducial-based ECG biometric detection methods do not require the detection of fiducial points. Instead, non-fiducial features are derived from segmented windows of ECG signals and may include autocorrelation coefficients and wavelet coefficients. Non-fiducial approaches often have a large number of redundant feature sets that need to be reduced [ 10 ]. Hybrid methods combine both fiducial and non-fiducial techniques by locating only the R-peaks. These are used to segment the ECG signal into single heartbeat waveforms, from which time or frequency domain information is extracted as features [ 8 ].

The final stage of the ECG recognition process consists of classification or pattern recognition. In this stage, the selected features of ECG signals are used as inputs for a classifier. The accuracy of the selection of the features will determine whether the classifier makes a correct or incorrect decision. There are many classification methods that have been proposed in recent years, including Bayesian Network, Linear Discriminant Analysis, Decision Trees, k-Nearest-Neighbors, Support Vector Machines, and Artificial Neural Networks. Each approach has its own advantages and disadvantages [ 10 ].

The success of identifying an individual through their ECG depends on the conditions they are exposed to during the acquisition process, the features that are extracted, and the classifiers used for identification or authentication. It is, therefore, important to evaluate the impact that certain changes have on biometric identification results [ 11 ]. This systematic review aims to discuss past research on the impact of variables in the data acquisition methods of an ECG signal on the biometric recognition process.

The paper is organized as follows: Section 2 presents the review methodology. Section 3 provides an overview of ECG acquisition and databases, which are discussed in Section 4 . Finally, the conclusions drawn are presented in Section 5 .

2. Review Methodology

In this section, the search strategy, which includes the identification, screening, and inclusion phases, is described, and the research questions we aim to answer are presented.

2.1. Search Strategy

This systematic review is structured according to Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [ 12 ]. Our literature research was performed from 5 June to 7 October 2022 in the Scopus database. The process of literature search for this literature review is illustrated in Figure 2 , and it is a three-stage process consisting of identification, screening, and inclusion.

Flow diagram of the literature research process (adapted from Prisma Guidelines [ 12 ]).

2.1.1. Identification

For the identification stage, the following general search terms were compiled for the Scopus research on the title, abstract, and keywords fields: (biometric* OR biometry AND ecg* OR electrocardiogram* OR electrocardiography* OR electrocardiographic* OR heart* AND authentication OR identification OR verification OR recognition AND “data collection” OR “signal collection” OR acquisition* OR collection OR signal* OR “body sensors” OR “body sensor” OR sensor* OR biosensor* OR database* OR electrode*). This search resulted in 958 papers. Six of those were duplicates and were consequently removed. Thus, the identification stage resulted in a total of 952 papers.

2.1.2. Screening

Before moving to the manual process of screening, we applied some exclusion criteria in our research. The first criterion concerned the year of publication of the article; only the ones published between 2000 and 2022 were considered. The second criterion was related to the subject area and the following areas were included: computer science, engineering, mathematics, materials science, and decision science. All remaining areas were excluded from our search. The document type was also an exclusion criteria: only conference papers, articles, and reviews the types of paper were considered. Lastly, only papers in English were included. A total of 137 articles were excluded based on the inclusion and exclusion criteria. A total of 815 papers were retrieved from the exclusion criteria process. The second part of the screening stage was a manual process of document exclusion. The purpose of this step was to filter the articles based on their abstract, methodology, results, or findings section to ensure that the articles match the goal of this systematic review. The screening process involved two rounds. In the first round, filtering, and screening were performed to exclude studies based on their respective title and abstract. Studies that did not focus on ECG-based biometric recognition were eliminated in this stage, and a total of 542 articles continued to the following round. The second round performed filtering by an accurate full-text reading of the examined articles from the first round based on an accurate full-text reading. Studies were eliminated based on the following exclusion criteria: (1) not focusing on the data collection process; (2) using ECG for non-biometric purposes; (3) not developing a biometric system algorithm; (4) not available online; and (5) using ECG combined with other biometric traits in a multimodal system. A total of 285 papers were eliminated due to reason (1), seventeen (17) due to reason (2), twenty-eight (28) due to reason (3), thirty-eight (38) due to reason (4), and fifty-three (53) due to reason (5).

2.1.3. Inclusion

After the screening process, 99 studies were integrated into our search. However, we also added some other reports from citation searching (7), resulting in a total of 106 studies included in this systematic review. The majority of papers included were from the journals/conferences presented in Figure 3 (top). The bottom of this figure presents the temporal increase of research on ECG-based biometric systems.

Main journals of publications of the papers included in this systematic review ( top ), and temporal increase of the research on ECG-based biometric systems ( bottom ).

2.2. Research Questions

This work mainly aims to provide some answers to the following questions about the ECG data for biometric systems:

Question 1: How are the ECG signals collected for biometric systems? What is the acquisition hardware information? This review compares the various aspects of the acquisition hardware, such as the intrusiveness of the acquisition (on-the-person vs. off-the-person acquisitions), the number of leads used, and the duration of the acquisition. We also present the most used commercially available and the self-developed acquisition devices and compare them in Table 1 .

Overview of the commercially available and self-developed devices.

Question 2: What should the acquisition protocol look like for a biometric system? Which conditions of acquisition should be considered? The aim of the present systematic review is to evaluate and compare the acquisition protocol of different research concerning the number of subjects and the assessment of the stability of the ECG signal over time. Moreover, since the health status of the subjects is also considered by many researchers, this systematic review presents literature findings regarding the impact of physical conditions, posture, emotions, and cardiac conditions on a biometric system.

Question 3: Which ECG datasets are used for biometric purposes? What are the main differences between them? These questions are addressed in Section 3.3 by presenting a description of the most used ECG databases in the literature and by providing a comparison between them in Table 2 .

Overview of the most used databases in the literature.

Question 4: Which factors of the data acquisition influence the intra- and inter-subject variability? What impact can these two variables have on the performance of a biometric system? The answers to these questions are discussed in Section 4 , in which the sources of intra and inter-subject variability are described.

3. ECG Acquisition and Databases

Regarding ECG acquisition, we covered the characteristics of the systems and protocols, as well as commercially available and self-developed devices. Finally, this section presents an overview of the databases used for biometric purposes.

3.1. Data Acquisition

Data acquisition can be organized according to criteria, such as the acquisition hardware information and acquisition protocol.

3.1.1. Acquisition Hardware Information

To analyze the characteristics of the acquisition systems, we should consider the intrusiveness of the acquisition and the types of electrodes, the number of leads used, and the duration of the acquisition.

Intrusiveness

Since the early research on ECG-based biometrics, the configurations used for data acquisition have significantly evolved. Researchers have mostly focused on addressing the main disadvantage of ECG as a biometric trait: its intrusiveness during data acquisition [ 5 ]. This has led to the development of off-the-person data acquisition methods which are less intrusive than traditional medical settings that use multiple wet electrodes. Data acquisition methods can be broadly divided into two categories based on their level of intrusiveness:

Acquisition methods that require attachment to the body, such as wet Ag/AgCl electrodes applied to the skin with a conductive electrolyte gel, are known as on-the-person methods. This approach relies on half-cell potential, double-layer capacitance, and parallel and series resistances to function. Despite providing good signal quality, wet electrodes can irritate the skin and restrict the user’s movement and may also cause interference between neighboring electrodes. These factors must be considered when using on-the-person data acquisition methods [ 43 ].

This type of acquisition can either be medical or unrestricted by movement, such as through Holter systems. According to medical standards and guidelines, the standard 12-lead configuration allows for the acquisition of an ECG signal in 12 leads (or channels) using three bipolar limb leads, three monopolar limb leads, and six monopolar precordial leads. The orthogonal configuration, also called Frank leads, allows the acquisition of ECG signals using seven electrodes. In early ECG biometric research, recordings from standard 12-lead and Frank leads were used in the development and evaluation of algorithms [ 44 , 45 , 46 , 47 , 48 , 49 , 50 ]. Some researchers chose acquisitions without movement restrictions, with longer duration, and with fewer electrodes, such as Holter systems, which can acquire ECG signals for several hours while subjects perform their daily activities [ 7 , 51 ].

The off-the-person acquisition method refers to devices that are integrated in objects or surfaces with which the subjects interact (e.g., a computer keyboard) and do not require any special preparation of the subject [ 52 ]. Wet electrodes characteristic of medical acquisitions were replaced by dry metallic electrodes, which increase the long-term performance and cause low skin irritation. However, they have high impedance between the electrode and skin, and are susceptible to motion artifacts [ 43 ]. Off-the-person acquisitions reduce the number of leads to two or three, and their placements are confined to the upper limbs, especially on the wrists, hands, or fingers [ 5 , 15 , 16 , 20 , 24 , 53 , 54 , 55 ]. Recently, a few initiatives have been conducted to improve off-the-person configurations and approach unconstrained settings in ECG biometrics. These efforts seek to close the gap to real, commercial applications by developing wearable technologies for ECG acquisition or embedding the sensors into common objects [ 7 , 11 , 13 , 18 , 28 ].

Table 3 presents a comparison of on-the-person and off-the-person acquisitions. Some researchers also compared different types of data acquisition and their influence on the performance of the biometric system. Jyotishi et al. [ 56 ] evaluated their model using three on-the-person ECG databases and two-off-the-person ECG databases. The results showed that the model performs well for both off-the-person and on-the-person ECG data. Srivastva et al. [ 2 ] used two databases, one on-the-person and the other off-the-person, and even mixed them together in a large database. The identification accuracies achieved for both the on-the-person and the off-the-person databases were individually about 99%, whereas an accuracy of approximately 98.5% was obtained for the mixed database. Thus, the authors proved the robustness of their ECG biometric method from signal acquisition methods. Biçakci et al. [ 7 ] used data from two different acquisition devices—the wearable-based chest bands and the medical-based Holter—to investigate whether the models are consistent and not biased by device specifications, providing reliable biometric verification with wearable devices. The results achieved for both datasets presented an equal error rate (EER) of around 5% for an enrollment time of 150 s, proving the reliability of using wearable devices for ECG acquisition for biometric purposes. It is also important to note that off-the-person methods have been gaining popularity in recent years for various applications beyond biometric recognition, such as disease detection. For example, in [ 57 ], the authors presented a method for recognizing diseases related to ECG and EEG data using sensors available in off-the-shelf mobile devices as well as sensors connected to a BITalino device. This suggests that these types of practical and convenient signal acquisition methods can be useful for a wide range of applications beyond biometric recognition.

Comparison of on-the-person and off-the-person acquisitions (adapted from [ 58 ]).

Number of Leads

A standard 12 lead ECG (or even 15-lead ECG) system can record more abnormalities than a single-lead ECG (similar to lead I in a 12 lead ECG). Figure 4 shows a representation of the standard 12-lead and the orthogonal-lead configurations. Due to the practical difficulty of collecting 12-lead ECG, biometric systems with a reduced number of leads have been evaluated.

Electrode placement and leads on the standard 12-lead configurations ( left ) and Frank leads ( right ), with the anterior electrodes depicted in red and the posterior electrodes depicted in a lighter red (adapted from [ 5 ], original figure kindly provided by the authors).

Dong et al. [ 44 ] proposed an identity recognition system and investigated their behavior on the different ECG leads. The experiments based on one-lead ECG showed that the best classification performance was obtained based on lead III and lead V1 and the worst classification performance was obtained based on lead V6. Moreover, experiments based on two-lead ECG outperformed experiments with one-lead ECG. Jekova et al. [ 59 ] used 12-lead resting ECG and evaluated the influence of the different leads on a biometric system. The capability of single limb leads was the lowest in III and aVR, and the highest in I and II. The identification capability of single chest leads was the lowest in V3 and the highest in V1. Multi-lead identification models yield considerably higher accuracy (about 20%) compared to the best single-leads. Porée et al. [ 47 ] proposed a method testing n = 1, 3, 6, and 12 leads, with all possible n-combinations of 12 leads tested. The best performances were obtained with 12 leads and then decreased with the decrease of the number of leads. With n = 3 and 6, the identification rate (IR) was still greater than 90%, whereas for n = 1, the IR was always lower than 90%. Fang et al. [ 60 ] tested their identification system with one- and three-lead ECGs, achieving optimal accuracies of 93% and 99%, respectively. Zhang et al. [ 9 ] suggested to place all the ECG electrodes on the left upper-arm or behind the ears in order to achieve excellent wearability. For the acquisition on the arm, the electrodes were integrated into an arm band, whereas for the acquisition behind the ears, electrodes were integrated into headsets or glasses. The signal strength of the single-arm ECG proved to be around 10% of the signal strength of the chest-ECG. However, arm-ECG heartbeats can still be distinguished. Moreover, the ear-ECG was found to be much weaker (5% of chest-ECG), but also shows a great potential for user identification purpose leveraging distinguishable morphologies. The mean accuracy obtained was as high as 98.8% and 91.1% for the single-arm and ear datasets, respectively.

Duration of Acquisition

Some researchers also assessed the impact of the duration of the ECG segment on the biometric identification’s performance. In the literature, it is predicted that the shorter the duration of the ECG segment used, the lower the performance obtained by the system. Ramos et al. [ 11 ] observed this behavior. However, when increasing the acquisition duration, this conclusion was not valid from a certain point onwards, as more data might introduce redundancy to the system. Results showed that around 10 s of signal are enough to test the identity of an individual. Biçakci et al. [ 7 ] evaluated the performance of their proposed method by varying the sample length. They used 5, 50, 150, 250, and 500 s of samples, achieving an EER of around 7% with only 5 s of enrollment. Ibtehaz et al. [ 61 ] studied the influence of the number of beats on the performance of the system, and results showed that increasing the number of beats significantly improved the performance. They achieved the perfect 100% accuracy using only three (ECG-ID [ 32 ], PTB [ 36 ] databases) and six (MIT-BIH Arrhythmia [ 34 ] and Normal Sinus Rhythm [ 35 ] databases) beats. Bernal-Romero et al. [ 62 ] tested their authentication method on different ECG signal duration: 10, 5, and 3 s. The EER rates for the authentication system with ECG signals had average values of 5.99%, 7.12%, and 9.66% for signal lengths of 10 s, 5 s, and 3 s, respectively. Pinto et al. [ 63 ] varied the number of enrollment templates between 5, 10, 15, and 30 s, achieving equal error rates of 13.70%, 10.92%, 9.52%, and 7.56%, respectively. Djelouat et al. [ 64 ] used testing times from 2 s to 5 s, and results showed an increasing identification rate from 84.44% to 98.88%. Carvalho et al. [ 23 ] aimed to measure the minimal number of heartbeats in which it was possible to identify subjects, even in situations where they were under the effect of fear or disgust, using from just one heartbeat up to twenty heartbeats. The results showed an optimal accuracy of around 75–80% when using 5–12 heartbeats.

3.1.2. Acquisition Protocol

Time stability.

The temporal separation between biometric evaluations may influence the system’s performance. Chee et al. [ 65 ] investigated the influence of different time separations between enrollment and testing data, using PTB [ 36 ] and ECG-ID [ 32 ] databases, with 83.9 days and 5.5 days between acquisitions, respectively. The model achieved accuracies of 64.16% and 92.70%, for long- and short-time separation, respectively, meaning that the model performance drops significantly when the time separation between the enrollment and classification increases. Ramos et al. [ 11 ] studied this influence by evaluating ECG segments from the same/different sessions. The results show that there is a decrease in performance when the sessions are different, except when the signal is collected on the fingers. Thus, the signal acquired on the fingers shows greater stability in the long term. Conversely, the point of acquisition that presents the greatest decrease in performance over time is the chest. Ibtehaz et al. [ 61 ] analyzed the cross-session accuracy using two databases, concluding that identification accuracy sharply falls when tested on data from a different session. Nevertheless, an accuracy above 90% was obtained for the ECG-ID database [ 32 ] and, for the CYBHi database [ 30 ], the accuracy dropped below 80%. Sun et al. [ 66 ] proposed a method for biometric identification, reaching a recognition accuracy of about 95%. However, they found that the accuracy degraded dramatically to 40% when considering a significant time interval between the acquisition of the training and testing templates. Porée et al. [ 47 ] proposed tests concerning the evaluation of the performances with time, concluding that there may exist a degradation of the ECG stability over time, with performances still acceptable after 16 months.

Number of Subjects

The impact of the database size on the accuracy of a biometric system is also a topic of research. Choi et al. [ 67 ] used 20, 40, 60, 80, and 100 subjects, reaching an accuracy of 100% for 20 subjects and an accuracy of 96% for 100 subjects, showing that the drop in accuracy when increasing the number of subjects is minimal. Jekova et al. [ 59 ] presented consistent validation of their identification models on an independent dataset by increasing its size from 10 to 230 subjects. Their validation results confirmed the expected trend for accuracy drop with the increase in the number of subjects. Chen et al. [ 53 ] evaluated their algorithm on 5, 10, 20, 30, 40, and 50 subjects. They noticed that the performance decreases slightly with the number of users increasing (false acceptance rate—FAR—increases from 0.00% to 8.00% and false reject rate—FRR—increases from 0.00% to 10.00% when increasing the number of subjects from 5 to 50). Carreiras et al. [ 17 ] tested the recognition system on an ECG signal database with 618 subjects. However, they also tested the system with subsets of this population, assessing the behavior of the recognition system with a varying number of subjects. For the entire database, results showed an EER of 9.01% and an Identification Error (IE) of 15.64%. The results of the population subsets highlight the fact that the EER does not seem to be affected by the population size, while, conversely, the IE increases with the number of subjects.

Physical Condition

Some researchers investigated the influence of exercise and body movement on the performance of a biometric system. Ramos et al. [ 11 ] studied the impact of the variability caused exclusively by moving the hands, feet, and chest. The results demonstrated that movement of the wrists causes the largest error in biometric identification, followed by movement of the fingers, while movement of the chest has almost no impact on the performance of the method. Huang et al. [ 68 ] showed that the ECG signal undergoes small noise interferences while the subject is walking and large noise when the subject is running or jumping. Nobunaga et al. [ 22 ] aimed to evaluate the effectiveness of their proposed identification method on exercising humans. They measured ECG during rest for one minute, with the subject lying down, and used these acquisitions to train the model. The exercise ECG used to test the model was measured for each subject after raising their foot so that their heart rate increased to over 100 bpm. The study reached an accuracy of 100% during rest and 99.8% during exercise, indicating that their method is accurate at identifying individuals doing exercise. Komeili et al. [ 54 ] also considered the case in which enrollment and testing are in different body conditions: rest and exercise. A feature selection was conducted to select features that are less affected by exercise; these were, then, used for enrolling and testing the biometric system’s users. Experimental results showed an EER of 11%. Moreover, Lee et al. [ 69 ] showed that the ECG cycle became shorter after 10 min of physical exercises, running, and holding breath for a certain period.

Most of the studies only consider supine rest conditions, which represent an important limitation regarding the use of ECG-based biometric systems in real-life contexts. Tirado-Martin et al. [ 21 ] acquired signals in different posture positions: sitting down at rest, standing at rest, and after exercise. They proved that different heart rates between the enrollment and recognition data result in lower performances. However, the best performance was achieved with the enrollment data acquired in a sitting position at rest. Iqbal et al. [ 70 ] achieved an accuracy of 100% when identifying 9 subjects at normal and resting conditions and an accuracy of 96.4% when identifying 39 subjects in 6 different physiological states (working, going up stairs, going down stairs, natural gait, lying with changed position and resting while watching TV). Wahabi et al. [ 71 ] considered an enrollment protocol in which each user’s ECG signal is collected under sit, stand, supine, and tripod postures. The accuracies achieved were 98.04% for sit, stand, and supine and 94.12% for tripod. Raj et al. [ 16 ] used ECG collected in three postures: a sitting posture at rest, a standing posture at rest, and a sitting posture after 20 s of exercise. They achieved an EER of 4.34% for the “standing” case, whereas the “sitting” and “after-exercise” cases worsened to 11.07% and 12.06%, respectively. Moreover, Wahabi et al. [ 72 ] also investigated the effect of body posture on ECG biometric accuracy, demonstrating that the performance of all the methods degraded when the train and testing data were not from the same body position. However, Porée et al. [ 47 ] showed that it is still possible to obtain good results even if the position in which the testing data was recorded is not present in the enrolment database.

An individual’s emotional state is continually changing. These changes occur naturally as a result of body chemistry, levels of stress, and even time of the day. The changes in emotional state are expressed in the ECG trace as changes in heart rate, noise in trace due to muscle flexor action, and variations in electrical potential gain. Thereby, some researchers have investigated the impact of emotions on the identification of individuals [ 73 ]. Zhou et al. [ 74 ] proposed a method of ECG biometrics using signals acquired under different stress levels, achieving an average recognition rate of 95%. Li et al. [ 18 ] used the public database DREAMER, in which the ECG signals acquired from wearable devices are disturbed by the physiological noises from emotional fluctuations induced by different stimuli. The accuracy obtained was of 91.30%, meaning that their method was capable of handling different kinds of emotional disturbances and identifying individuals accurately. Zheng et al. [ 75 ] investigated whether ECG based identification was affected by the status of ECG signal collecting, considering four status pairs: emotional status (calm, high pressure), eating (starve, satiation), sleeping (full, lack), and health (healthy, tired/cold). From the obtained results, the authors concluded that negative emotions (high pressure) and lack of sleep reduced the True Positive Rate (TPR) slightly (around 2–3%), but there was no effect on both eating and health pair status. When using ECG data mixed with all four statuses, the overall TPR of identification reached approximately 85%. Zheng et al. [ 76 ] self-collected calm and high-pressure ECG datasets to investigate the influence of different emotional statuses. They achieved accuracies of 98.10% and 95.67% for calm and high-pressure data, respectively, showing that the ECG signals under different emotion statuses can be used in reliable and accurate biometric systems. Israel et al. [ 73 ] used ECG data collected during seven different tasks performed to stimulate different states of anxiety. The low stress tasks were the subject’s baseline state, mediative, and recovery tasks. The high stress tasks were reading aloud, mathematical manipulation, and driving in virtual reality. Results showed that both within and between anxiety states, nearly all the individuals were correctly classified, as the accuracies obtained were around 97–98%.

Cardiac Conditions

The behavior of a biometric system under heart conditions has also been assessed. Chen et al. [ 77 ] focused on the comparative performance analysis of human identification with ECG signals collected from subjects in different health conditions. Data used consisted of ECG signals from 38 elderly subjects with a variety of chronic diseases and 30 young healthy students. Experimental results indicated that a better recognition accuracy is achieved for healthy subjects (98.14%) when compared to elderly unhealthy subjects (95.62%). Becerra et al. [ 78 ] used a database comprising 20 healthy subjects and 20 pathological subjects (diagnosed with different types of cardiac murmurs). The accuracies obtained were 91.19% and 97.74% for patients with cardiac murmurs and healthy patients, respectively. Singh et al. [ 79 ] used the QT database for patients with cardiac diseases and a second database for healthy patients. The proposed ECG biometric method achieved EER of 0.76% and 0.71% in recognizing people suffering from cardiac arrhythmia and people of good health, respectively. Regarding mixed health status, the method achieved an EER of 1.31%, confirming a very good performance and robustness of the proposal. Singh et al. [ 48 ] proposed a method to identify arrhythmic and normal subjects, reaching an accuracy of 87.37% for the subjects of MIT-BIH Arrhythmia database [ 34 ] and 92.88% for the IIT (BHU) database. Sidek et al. [ 80 ] also used three different databases containing various irregular heart states: MIT-BIH Arrhythmia database [ 34 ], MIT-BIH supraventricular arrhythmia, and Charles Sturt diabetes complication screening initiative, achieving accuracies of 96.7%, 96.4%, and 99.3% for each, respectively. Loong et al. [ 81 ] showed that diseased ECG only reduced the recognition rate by less than 1% and, thus, the system is robust towards diseased ECG. Contrarily, Chiu et al. [ 82 ] registered a drop of 19% between identifying normal subjects and subjects with arrhythmia (100% and 81%, respectively). Moreover, Ghazarian et al. [ 83 ] assessed the accuracy of ECG-based identification for distinct heart condition groups. They discovered that, in contrast to the initial expectation that identification accuracy for healthy normal sinus rhythm should be the highest, the identification accuracy is higher for patients with sinus tachycardia or patients who are diagnosed with both ST changes and supraventricular tachycardia. Conversely, they observed that patients with premature ventricular contractions have an identification accuracy as low as 78.54% and patients with a pacemaker presented an accuracy of 80.2%.

The conditions under which ECG data are acquired can have a major impact not only on the performance of a biometric system but also on the ability to accurately and reliably detect heart conditions. As such, several studies have been investigating the influence of several factors, such as electrode placement, lead configuration, physical exercise, and the intrusiveness of acquisition, on the detection of heart diseases [ 84 , 85 , 86 ].

3.2. Acquisition Devices

3.2.1. commercially available devices.

During the last years, the market of medical-grade wearable ECG devices has expanded, and these have increasingly been used for biometric purposes since they can be easily integrated in biometric systems, reducing costs, power consumption, and time of acquisition. However, researchers still use non-wearable devices for data acquisition since they allow records with higher quality. In this section, some of the most used commercially available acquisition devices are presented, as well as some self-developed sensors for ECG acquisitions. Figure 5 shows the commercially available devices described in this section.

Commercially available devices.

The VitalJacket [ 87 ], presented in Figure 5 a, is a wearable device developed by researchers from the IEETA research unit at the University of Aveiro and commercialized by Biodevices SA [ 88 ]. It is designed to continuously record high-quality ECG and other vital signals in various clinical and everyday settings. The collected data can be stored on an SD card for offline analysis or transmitted via Bluetooth to mobile devices for real-time monitoring and online processing. Ye et al. [ 13 ] investigated the applicability of ECG signals from such wearable device in human identification. In the five-subject study, their proposed method exhibited near 100% recognition rates based on single heartbeats, even with a six-month interval between the training and testing data. Ramos et al. [ 11 ] used VitalJacket to collect ECG signals from twenty healthy participants in two sessions separated by 2 weeks. They investigated the impact of movement, the influence of using different ECG acquisition placement, the impact of temporal separation between sessions, and the impact of the acquisition time. The authors reached an accuracy of 99% for signals collected on the fingers in two different sessions. For the various experiments, the results suggested that the ECG signals acquired using VitalJacket can be used as robust biometrics.

The RespiBAN Professional, which is presented in Figure 5 b, is a wearable system made by PLUX, which includes a PLUX accelerometer biosensor and biosignal acquisition hardware, as well as a respiration biometric sensor embedded in the chest strap fabric. This device can measure various biosignals, including ECG, electrodermal activity (EDA), electromyogram (EMG), and skin temperature. The collected data can be transmitted to mobile devices via Bluetooth. Biçakci et al. [ 7 ] used the WESAD dataset, which consists of ECG recordings collected from a RespiBAN device. The EER obtained was 7.07%, meaning that ECG biometrics will be a valid verification option (or could be in the future) using wearable devices for data acquisition.

The Nymi Band [ 89 ], shown in Figure 5 c, is a wearable device that uses the wearer’s unique cardiac signal to unlock Bluetooth-enabled devices such as computers, smartphones, and cards. To authenticate the user, the Nymi Band is placed on the wrists and the top of the device is touched with a finger. As long as the device is worn, the user remains authenticated. The Nymi Band is equipped with a heart rate monitor, accelerometer, gyroscope, and biometric authenticator, and is powered by a rechargeable battery. Chun et al. [ 14 ] used ECG data from 15 subjects collected using the Nymi Band, achieving an EER of 0.9%, which proves the reliability of this wearable device.

ReadMyHeart [ 90 ], shown in Figure 5 d, is a handheld, non-invasive heart monitoring device made by DailyCare BioMedical Inc. It allows users to record electrical signals from their hearts by placing their thumbs on the device’s conducting plates, without the need for wires or conducting gel. The device takes 30 s to record each measurement and displays the average heart rate, ST segment, and QRS interval. These readings are based on a “modified Lead I-ECG”, rather than traditional standard ECG readings. Islam et al. [ 15 ] captured ECG signals from 112 individuals using the handheld ECG device ReadMyHeart, achieving a minimum EER of 10.52%.

The Vernier ECG Sensor [ 91 ], presented in Figure 5 e, is a device used to measure the electrical potential waveforms produced during the contraction of the heart. It can be used to record standard three-lead ECG tracings or surface EMG recordings of muscle contractions in various parts of the body. The device is usually associated with the Vernier Go!Link interface, which is a low-cost USB sensor interface that connects Vernier sensors to a computer. Raj et al. [ 16 ] used the Vernier sensor for ECG acquisitions on the arm with different body postures, achieving an Optimal performance with an EER of 4.34%.

The PageWriter Trim III [ 92 ] is a compact and cost-effective cardiograph made by Philips, shown in Figure 5 f. It is an interpretative ECG system designed for fast-paced clinical environments, with features such as a high-resolution full-color display and the ability to report, store, and transmit 12-lead ECG data using industry-standard XML. Carreiras et al. [ 17 ] used Philips PageWriter Trim III for ECG acquisitions of 618 subjects, achieving an EER and an IE of 9.01% and 15.64%, respectively.

The Shimmer ECG unit [ 93 ] is a device designed for the measurement of physiological signals for ECG, and it is presented in Figure 5 g. It includes a configurable digital front-end and an ECG sensor that can record the pathway of electrical impulses through the heart muscle. The sensor can be used to record ECG data on resting and ambulatory subjects, or during exercise to provide information on the heart’s response to physical exertion. Li et al. [ 18 ] used a public database, DREAMER, in which data were acquired using the Shimmer ECG Sensor, and an accuracy of 97.2% was obtained.

The BioPLUX [ 94 ] low-noise ECG local differential triode configuration enables fast application and unobtrusive single-lead ECG data acquisition. This sensor can be used to extract heart rate data and other ECG features, enabling its application in research fields such as biomedical, biofeedback, psychophysiology, and sports, among many others. Silva et al. [ 19 ] used the BioPlux Electrocardiography Sensor presented in Figure 5 h integrated on a steering wheel for in-vehicle driver recognition, achieving an IE of 2.40%.

Maxim 86150 Evaluation Kit [ 95 ], presented in Figure 5 i, is a device designed to evaluate the photoplethysmogram (PPG) and ECG bio-sensor module. The device includes a Microcontroller Board and a Sensor Board. The Microcontroller Board houses a microcontroller unit (MCU) with preloaded firmware, Bluetooth connectivity, and power management. The Sensor Board includes the MAX86150 Bio-Sensor Module and two stainless steel dry electrodes for ECG measurement. The Evaluation Kit is powered by an included lithium polymer battery, which is charged with a micro-USB cable. When monitoring is active, the module uses IR Proximity Mode to detect each user’s fingers, and a red LED will turn on when a finger is near the module. Sorvillo et al. [ 20 ] used the Maxim 86150 Evaluation Kit to collect ECG for human identification under rest and mental and physical stress, reaching accuracies of 88% and 68%, respectively.

The BioRadio [ 96 ] is a wireless biomedical monitor, shown in Figure 5 j, with programmable channels for recording and transmitting various combinations of human physiological signals. It is easy to set up and operate, and the wearable device captures data in a flexible file format compatible with a variety of software suites and proprietary tools. Huang et al. [ 68 ] used the BioRadio device with the positions of the electrodes following the Einthoven’s configuration. Abdelazez et al. [ 97 ] also used this device, but the electrodes were positioned under the right and left thumbs instead. Their system achieved a precision of 0.68, being able to identify 98.7% of the false positives while retaining the true positives rate.

The BIOPAC MP160 [ 98 ] is a 16-channel system designed for the acquisition of various physiological signals, including Heart Rate Variability (HRV), Electroencephalogram (EEG), EMG, EGG, and many more. The device, represented in Figure 5 k, offers multiple configurations to suit different research and teaching needs, and records multiple channels with different sample rates up to 400 kHz. Used in conjunction with AcqKnowledge software and BIOPAC electrodes, amplifiers, transducers, and other system components, the MP160 is part of a complete data acquisition and analysis system. Many researchers used the BIOPAC system for data acquisition of their proposed biometric system [ 21 , 22 , 23 ].

Kardia [ 99 ] is a wireless device that allows users to record a medical-grade single-lead ECG in 30 s and receive instant analysis on their phones. It is clinically validated, CE marked, and FDA-cleared, making it a reliable option for checking one’s heart from home. Kardia, in Figure 5 l, is compatible with most popular phones and tablets and, to use it, one only needs to download the Kardia app. Arteaga-Falconi et al. [ 24 ] used the Kardia device along with a mobile phone for ECG acquisitions at different times and conditions, proving the reliability of this mobile device, since the results revealed 1.41% of FAR and 81.82% of true acceptance rate.

3.2.2. Self-Developed Acquisition Devices

As mentioned above, there are also some researchers who developed their own ECG sensors.

Rashkovska et al. [ 25 ] developed a wireless ECG sensor for long-term monitoring and tested it in various applications, including biometric authentication. The initial prototype of the wireless body sensor (WBS) was powered by a coin-sized battery and included a low power microcontroller and a 2.4 GHz radio transceiver. The design was later improved to include a rechargeable battery and a Bluetooth Low Power (BLE) radio transceiver for communication. The WBS is attached to the skin by using self-adhesive electrodes and has evolved into a more flexible and lightweight design that allows for unobtrusive long-term health monitoring and low-cost implementation. It is now commercially available as the SavvyTM sensor. The proposed methodology for biometric authentication using this device achieved an EER from 6% to 13%, depending on the subject.

Blasco et al. [ 26 ] developed a wearable sensor capable of measuring photoplethysmography (PPG), ECG, Galvanic Skin Response (GSR), and Acceleration (ACC) signals from the wrists. The ECG sensor is from Bitalino [ 100 ] and the two electrodes were placed on the inner side of the wristband and on top of the wristband, respectively. The viability of the use of the sensor on a biometric system was tested in three different acquisition conditions: sitting, walking, and sitting after exercise, and the results were promising.

Guven et al. [ 3 ] also developed a fingertip ECG data acquisition device for biometric purposes. The device consists of two dry-contact sensors, produced by Plessey Semiconductors, an instrumentation amplifier, an anti-aliasing filter, an optocoupler, a digital signals controller (DSC), and a USB connection unit. The authors conducted an experiment to evaluate the performance of the proposed device by comparing it to the use of lead-I ECG signal, recorded using Biopac MP36 with three conventional Ag/AgCl electrodes and gel. The results achieved were around 100% for the IE, showing that this portable, inexpensive, and user-friendly device is very promising for biometric applications.

Wieclaw et al. [ 10 ] developed a sensor using an Arduino Uno and e-Health Sensor Platform V2.0 for data acquisition. Arduino Uno is a microcontroller board with 16 MHz quartz crystal and a USB port for programming, debugging, and data transfer. The e-Health Sensor Platform V2.0 extends the Arduino Uno and enables the implementation of biometric and medical applications. Data acquisition was performed using differential OpAmp schema followed by 8-bit ADC operating at 277 Hz sampling rate. ADC data were transferred to a PC via the COM-port using the PySerial Library. Modified schema required the user to touch the electrodes with two fingers from the left hand and one finger from the right.

Peter et al. [ 27 ] used a low-cost sensor and designed a sensor processing board. They used conventional wet cloth electrodes with repositionable conductive adhesive hydrogel to measure the electrical activity from the skin surface. Then, a sensor board that amplifies and filters the signals was designed. They applied a standard difference amplifier approach, which is a suitable solution since the basic ECG data is obtained as an output of the difference of two leads placed on the body. The circuit build consists of three parts: the differential amplifier, a filter, and a post amplifier. Afterwards, the signal follows to an embedded target platform, called Raspberry Pi (RPi), which is a low-power single-board computer. One advantage of the RPi is that it is supported by MATLAB Simulink and a range of design tools, which facilitates easy and fast prototyping.

Ramli et al. [ 28 ] developed a portable ECG detection kit integrated into a wearable bracelet that is responsible for detecting the heartbeat signal of the user and sending out the ECG signals to be processed via Bluetooth. The sensor is equipped with three electrodes and by placing a finger on the topside electrode while the user’s wrists are in contact with the other two electrodes, an electrical circuit is completed; ECG signals are able to be detected by the device. The heartbeat detection kit is formed by six main parts: instrumentation amplifier (IA), high-pass filter (HPF), 60 Hz notch filter, low-pass filter (LPF), analog-to-digital converter (ADC), and signal transmitter. They also developed an Android platform application that acts as a secure login system. This application receives the serial data from the heartbeat detection kit through a Bluetooth connection. Then, when the sign-in or sign-up function of the application is triggered, the incoming data is saved to the database. The database will trigger the back-end system which is the Intel platform board to perform the embedding, features extraction, and pattern-matching processes. Once the processes are done, the verification result will be sent to the android application GUI.

Lourenço et al. [ 29 ] proposed a method and device for ECG acquisition, using a single lead setup at the fingers, with Ag/AgCl electrodes without gel. This setup aims to increase the usability and acceptability of ECG-based biometric systems to the level of other biometric traits in terms of signal acquisition. The rigid base integrates three leads which, due to the underlying sensor design, correspond to the ground, positive, and negative poles. The right-hand thumb is used as a negative electrode and the left-hand index finger acts simultaneously as the positive and ground electrodes. The base sensor is an ecgPLUX active ECG triode and the transmission was done via a Bluetooth wireless bioPLUX research biosignal acquisition unit.

Table 1 show an overview of the commercially available and self-developed sensors described in Section 3.2.1 and Section 3.2.2 .

3.3. Databases

Currently, there are several collections publicly available for ECG biometrics research. Below, the most relevant of the currently available ECG collection are characterized. Table 2 summarizes the characteristics of each. Some are publicly available and can be found on physionet [ 101 ].

3.3.1. On-the-Person

The MIT-BIH Arrhythmia database [ 34 ] is a widely used resource for ECG-based biometrics research and is available at the Physionet repository. It consists of 48 half-hour ECG recordings from 47 subjects that were collected in the laboratories at Boston’s Beth Israel Hospital. Out of 27 subjects, 23 recordings were selected from a mixed population of inpatients (about 60%) and outpatients (about 40%), and the remaining 25 recordings were selected from the same set to include less common but clinically significant arrhythmias. The recordings were digitized at 360 samples per second per channel with an 11-bit resolution over a 10 mV range [ 49 ].

This database is composed of excerpts from 18 subjects from the MIT-BIH Arrhythmia database presented above, which are deemed to be free from arrhythmias or other diseases. Subjects included in this database were found to have had no significant arrhythmias, and they include 5 men, aged 26 to 45 and 13 women, aged 20 to 50 [ 35 ].

This database contains 25 long-term ECG recordings of human subjects with atrial fibrillation (mostly paroxysmal). The individual recordings are each 10 h in duration and include two ECG signals. The original analog recordings were made at Boston’s Beth Israel Hospital using ambulatory ECG recorders with a typical recording bandwidth of approximately 0.1 Hz to 40 Hz [ 40 ].

This database is obtained by the Physikalisch-Technische Bundesanstalt (PTB), National Metrology Institute of Germany [ 36 ]. The database contains 549 records with diverse profile information and various lengths of ECG from 290 subjects. Of the 290 subjects, 148 had suffered from myocardial infarction, 18 had cardiomyopathy or heart failure, and 52 were healthy subjects. Acquisitions were performed both through the standard 12-leads and the three Frank leads [ 49 ].

The ECG identification database was recorded for biometric identification purposes [ 32 ]. Each raw ECG record was acquired for about 20 s with a sampling rate of 500 Hz and a 12-bit resolution. The first two records acquired on the same day were used for each subject. The database consists of 310 one-lead ECG recording sessions obtained from 90 volunteers during a resting state. The number of sessions for each volunteer varied from 2 to 20, with a time span of 1 day to 6 months between the initial and last recordings [ 49 ].

This is an ECG database from the University of Rochester that is focused on biometrics. The study population consists of 202 healthy subjects from the Intercity Digital Electrocardiogram Alliance (IDEAL) database. The database includes 24 Holter recordings that were acquired using the SpaceLab-Burdick digital Holter recorder. The equipment provides 200 Hz sampling frequency signals with 16-bit amplitude resolution. The ECG was acquired using a pseudo-orthogonal lead configuration (X, Y, and Z), obtained through four electrodes placed on the chest. There is an initial resting supine period with a duration of 20 min before starting the ambulatory recording [ 33 ].

The QT database is a collection of ECGs that have been selected to showcase a wide range of QRS and ST-T shapes, with the goal of testing QT detection algorithms with real-world variability. These records were largely drawn from various ECG databases, including the MIT-BIH Arrhythmia Database [ 34 ], as well as additional recordings gathered at Boston’s Beth Israel Deaconess Medical Center. The additional recordings were chosen to represent extreme examples of cardiac (patho)physiology, including data from Holter recordings of patients who experienced sudden cardiac death during the recordings, as well as age- and gender-matched patients without diagnosed cardiac disease. The QT database includes a total of 105 fifteen-minute excerpts of two-channel ECGs [ 37 ].

This database contains data collected from a real-world driving task designed to measure a driver’s stress level. The driving protocol involved following a predetermined route for 20 min on open roads in the Boston area while following a set of instructions. Four types of physiological sensors were used during the experiment: ECG, electromyogram (EMG), skin conductivity (EDA and GSR), and respiration. These sensors were connected to a FlexComp analog-to-digital converter, which isolated the subject from the power supply. The ECG electrodes were positioned in a modified lead II configuration to minimize motion artifacts and maximize the amplitude of the R-waves, and the ECG was sampled at 496 Hz. In total, 27 collections were recorded, 6 from drivers who completed the course only once, and 7 from 3 drivers who repeated the course on multiple days [ 31 ].

The Fantasia Database is a collection of 120 min of continuous ECG recordings taken while subjects were lying down. Two groups of healthy human subjects, ten young and ten elderly participated in this acquisition. Only healthy, nonsmoking subjects with normal exercise tolerance tests, no medical problems, and taking no medication, were admitted to the study. The subjects laids supine for 120 min while continuous ECG signals were collected. All subjects remained in a relaxed state with a normal sinus rhythm while watching the movie “Fantasia” from Disney to help maintain wakefulness [ 39 ].

3.3.2. Off-the-Person

In this work, Silva et al. [ 30 ] presented the CYBHI database which consists of 128 ECG recordings acquired using the off-the-person approach. The ECG signals (2 min long) were recorded simultaneously from both wrists and fingers using dry Ag/AgCl electrodes and electrolycra strips, respectively. These sensors were placed on custom hand-shaped support, and data synchronization was ensured using the syncPLUX synchronization kit. The electrodermal activity data was also collected to provide information about the arousal state of the subject, as the acquisition protocol included both neutral and emotional elicitation tasks. The acquisition protocol consisted of short-term and long-term sessions. Short-term sessions were conducted over 2 days with 65 participants. The participants completed an experimental procedure that was 5 min long, during which they watched a low-arousal video and a high-arousal video (a horror movie trailer). Long-term sessions consisted of 2 data acquisition moments separated by a 3-month period with 63 participants. In both phases, only ECG signals from the fingers were recorded, and in each of the sessions, the subjects were seated for 2 min in a resting position with two fingers on the dry Ag/AgCl electrodes [ 52 ].

Pouryayevali et al. [ 38 ] collected a large database with 1012 ECG recordings from different people. The acquisition hardware consisted of a pad with dry Ag/AgCl electrodes, positioned so that the left thumb was placed on the positive electrode, whereas the right thumb and right forefinger were placed on the negative and reference electrodes, respectively. According to the acquisition protocol, the ECG recordings were performed in the following conditions: supine, tripod, sit, physical exercise, and stand. The ECG signals were recorded for all the subjects while sitting, but they were collected in supine, tripod, physical exercise, and standing conditions only for 63, 63, 71, and 81 participants, respectively. Regarding the time interval, 72, 65, 54, 47, and 43 out of 1012 subjects participated in 2, 3, 4, 5, and 6 acquisition sessions, respectively. The length of each recording ranged from 2 min to 5 min [ 52 ].

This database contains two-lead ECG recordings taken during affect elicitation using audio-visual stimuli. The data was collected using eight film clips containing scenes from different films that were designed to elicit a range of emotions. Of these eighteen film clips, two were intended to evoke the following nine emotions: amusement, excitement, happiness, calmness, anger, disgust, fear, sadness, and surprise. The film clips were between 65 and 393 s long. ECG was recorded using a SHIMMER wireless sensor. A total of 25 healthy volunteers aged between 22 and 33 years old participated in the study [ 41 ].

This dataset consists of ECG recordings, along with several other physiological signals, collected from 15 subjects using a RespiBan device. The device was placed around the subject’s chest and recorded ECG using a standard three-lead configuration. After the subjects were equipped with the sensors, a 20-min baseline was recorded (baseline condition), while sitting/standing at a table. Neutral reading material (magazines) was provided. During the amusement condition, the subjects watched a set of eleven funny video clips. The following phase was a stress condition, in which the subjects were exposed to the Trier Social Stress Test (TSST), which consists of a public speaking and mental arithmetic task. The subjects had to deliver a five-minute speech on their personal traits in front of a panel of three people, focusing on strengths and weaknesses. After the speech, the panel asked to subjects to count down from 2023 to 0, in steps of 17, and asked to start over if they made a mistake. Both tasks lasted about 5 min, resulting in a total of ten minutes for the stress condition. The amusement and stress conditions were followed by a meditation, in which subjects followed instructions with their eyes closed while sitting in a comfortable position for seven minutes [ 42 ].

4. Discussion

Based on the data reported so far, the following issues are discussed: (i) the comparison of the acquisition hardware, (ii) the comparison of the acquisition protocol, (iii) inter-subject variability, and (iv) intra-subjects variability.

4.1. Acquisition Hardware

4.1.1. acquisition devices.

Most databases were collected from medical devices that often have more leads, which makes them more informative. However, the large number of electrodes required, their uncomfortable placement, the limited movement allowed, and the duration of recordings, make it difficult to develop robust biometric systems. Some researchers have attempted to address these issues by using acquisition methods that allow more movement and longer durations using fewer electrodes. One of the most prominent examples was the use of Holter systems, which are designed to continuously acquire ECG signals for several hours while the subjects move and perform daily activities. The Holter monitors are smaller devices compared to standard 12-lead ECG devices, but they still use many wires to connect the electrodes to the recording machine. While these monitors have become more advanced and capable of recording high-quality single or multi-lead ECG, they can still be uncomfortable for patients to wear and can affect the ECG signal’s strength as a biometric trait [ 25 , 102 ].

Despite the potential of off-the-person systems in a practical setting, there are still some challenges that need to be overcome. Off-the-person systems still require the user to hold the electrode or deliberately place the fingers or palms over them. This prevents us to designate them as unconstrained systems, which puts the ECG at a disadvantage over other biometric traits that can already be used for unconstrained recognition. In addition, the use of dry electrodes in farther placement makes the acquisition more vulnerable to interferences, thus affecting the quality of the signal. The efforts on wearable devices have brought ECG biometrics closer to viable, unconstrained applications.

However, the adoption of wearables for biometrics also introduces new challenges [ 26 ]. First, wearable devices tend to use cheaper sensors and hardware than traditional biometric systems. Consequently, sensor readings have more noise, and combined with natural variability in the subject’s state, accuracy is more of an issue. Furthermore, wearable devices have limited computational capabilities and must optimize the usage of their resources to maximize battery life, while providing a quick response to biometric challenges. Hence, these issues must be addressed and adequately solved in order to obtain viable commercial ECG biometric systems [ 3 ].

4.1.2. Number of Leads

It is known that each ECG lead contains discriminative information regarding the electrical activity of the heart. However, the use of 12-lead ECG is a very unpractical solution for biometric purposes, as it requires the placement of many electrodes. Hence, biometric systems have been evaluated with a reduced number of leads, since it represents interest from a technological point of view. Some researchers, such as Porée et al. [ 47 ], tested the use of different combinations of leads (n = 1, 3, 6, and 12 leads) to evaluate the behavior of the system. As expected, they achieved optimal performances with 12 leads and then the performance decreased with the number of leads. However, with only one lead, the performance was still close to 90%. Concerning the use of single-lead ECG, researchers like Jekova et al. [ 59 ] used all 12 leads individually as single-lead configurations in order to assess the influence of each lead. They stated that the capability of single limb leads is highest in lead I (and lead II), which justifies the use of lead I in almost all studies in the literature. Moreover, Zhang et al. [ 9 ] tried to reach high wearability by placing the electrodes on the upper arm or behind the ears, using an armband and headsets, respectively, proving that, although the strength of the ECG is much lower than the chest-ECG, it still has a great potential for user identification purposes. Thus, more efforts should continue to be made to improve performance while using a single-lead ECG, gathered with minimal intrusiveness, as it is a much more user-friendly approach.

4.1.3. Duration of Acquisition

Short-term ECG data (less than several minutes) and long-term ECG data can complement each other. Short-term ECG data is cheaper and easier to collect. Many cardiac diseases can be detected based on short-term ECG, so such data represent the primary diagnostic tool in outpatient departments. However, long-term ECG can help to detect diseases with intermittent symptoms such as paroxysmal ventricular fibrillation (VF) and atrial fibrillation (AF) [ 103 ]. Data acquisition should be a relatively fast process for biometric scenarios. Nevertheless, it is predicted that the shorter the duration of the ECG segment used, the lower the performance obtained by the recognition system. Thereby, many studies assessed the impact of the duration of the ECG segment on the performance of the biometric system, as mentioned above. In general, this behavior was observed in most studies. However, Ramos et al. [ 11 ] showed that this pattern may not always be valid from a certain point onwards, as more data can introduce redundancy to the system. While optimal performances were achieved with ECG segments of 10 s in [ 11 ] and [ 62 ], Djelouat et al. [ 64 ] reached an accuracy of 96.66% with only 5 s of acquisition. Thus, it can be suggested that the optimal duration of acquisition may depend on the conditions of acquisition. Taking into account that the duration of the acquisition for a biometric system should be short, the optimally short acquisition time that does not compromise the performance of the system should be investigated.

4.1.4. Sampling Frequency of Acquisition

In general, to ensure that ECG signals are recorded with sufficient detail and resolution, a sampling frequency of at least 500 Hz is commonly used, since it will allow capturing fast changes in ECG signals [ 104 ]. When the sampling rate is lower, more information can be lost in the recording and there is a greater change of high-frequency noise being misinterpreted [ 5 ]. According to [ 105 ], due to the lack of points available in ECG signals, signals with low sampling frequency are usually inefficient for ECG matching purposes when methods such as cross-correlation, percentage root-mean-square deviation, and wavelet distance measurement are used. As such, the low sampling frequencies used in old commercial systems (e.g., 128 Hz) usually need a reinterpolation of data [ 106 ]. There are already some studies that developed enhancement techniques to increase the number of samples of a given ECG data from low sampling frequency recordings [ 107 ]. Nevertheless, the frequency at which the ECG should be recorded depends on the specific application and the equipment used for recording [ 104 ].

In the past, waves with smaller amplitudes were usually ignored since these were almost always caused by noise. However, with the advent of high-resolution ECG technology, it became possible to detect signals as small as 1 μ V through the use of signal averaging techniques. This advancement has provided new insights and has shown that techniques to reduce noise are effective in clinical settings. In [ 106 ], the authors investigated the effects of the amplitude resolution of ECG acquisition systems on the P-wave analysis. Results proved that at lower bit resolutions, the percentage error is higher than 40%, meaning that almost half of the results are different from the ones obtained with the highest resolution. They concluded that the high resolution used in modern electrocardiographs was suitable for ECG analysis.

There is no standard regarding data acquisition hardware information. Different studies have used various sources, number of leads, and durations of acquisition, making it difficult to compare results between different datasets. However, by analyzing the literature, it can be concluded that the preferable solutions for the acquisition hardware of a biometric system are the following: (1) an off-the-person approach, which is more realistic to a biometric scenario and easily integrated into a biometric system; (2) a one-lead setup, as it is the most practical solution due to the reduction of the number of contact points, making the data acquisition of benefit to user acceptance; (3) a short-term ECG data, since such data would be faster to acquire and to process, resulting in a user-friendly biometric system.

4.2. Acquisition Protocol

4.2.1. number of subjects.

There is an expected trend for a performance drop with an increase in the number of subjects; several studies have confirmed this expectation by testing the system with a different number of subjects and registering a decrease in performance when increasing the number of subjects [ 53 , 59 ]. However, a biometric system should be able to accurately identify and authenticate many subjects. Although the vast majority of studies have tested their models with small databases, Carreiras et al. [ 52 ], for example, used a database with 618 subjects, achieving promising results both on authentication and identification (EER of 9.01% and EI of 15.4%). Thus, larger databases, as well as subsets of those databases, should be used in biometric systems to assess their behavior according to the number of subjects considered.

4.2.2. Time Stability

Multi-session ECG authentication, with enrollment and authentication signals captured across two or more different sessions, has become a more relevant problem of late, primarily because of its similarity to real-world use cases. Error rates calculated using multi-session datasets have been reported to be significantly worse than those using single sessions. Two acquisition sessions in a dataset may differ in several different ways, e.g., signals may be captured under different postures or heart-rate, different hydration levels, or may be captured on different days [ 108 ]. Performance degradation might arise from the morphological changes in the heartbeats from one session to another due to variations in physical or physiological states of these subjects [ 45 ]. Ramos et al. [ 11 ] found a decrease in the performance over time, except when the signal is collected on the fingers, with the chest-ECG being the one that obtained the greatest decrease in performance over time. Thus, off-the-person approaches, which are the most user-friendly, may not suffer significant degradation over time.

4.3. Sources of Variability

The ECG signals could be affected by three major sources of variability, namely inter-subject and intra-subject variability and artefact and noise, which will be individually described.

4.3.1. Inter-Subject Variability

Inter-subject variability is the variability between ECGs from different individuals. The ECG signal uniqueness can be assumed to be acquired mainly from the uniqueness of DNA, besides other physical factors such as age, ethnicity, and gender, which contribute to the different ECG variations. Since the ECG signal is universal, stable, and unique, the inter-subject variability can be affected by the orientation of heart mass and the conductivity of cardiac muscle. Despite this inter-subject variability, the ECG signal should remain sufficiently stable over time to enable ECG-based biometric authentication. The main reasons for inter-subject variability of ECG morphology are the heart geometry and the individual attributes.

• Heart Geometry: Heart size, cardiac muscle thickness, and the overall shape of the heart dictate the paths the electrical current follows inside the heart, the number of muscle cells that depolarize, and the time it takes to depolarize the whole heart. Athletes, due to their high levels of physical training, commonly have larger hearts with thicker myocardia, which affects the ECG with higher voltages in the QRS complex and results in lower basal heart rates [ 109 , 110 ].
• Individual Attributes: Age, weight, and pregnancy are some of the individual attributes that can cause shifts in the heart position and/or orientation. These shifts will change the orientation of the electrical current conduction vectors along the heart, meaning the electrodes will detect the signal from a different perspective, thus altering the ECG waveform. For instance, the QRS complex amplitude tends to increase from birth to adolescence and then gradually begins to decrease afterward [ 8 ]. The authors in [ 111 ] also find that the PR interval increases slightly with increasing age. Studies have shown that the amplitude of the S wave in ECG signals is lower in women than in men between the age interval of 18–40 [ 112 ]. While gender differences in ECG signal parameters are more evident in young adulthood, they are known to decrease their effect afterward.

4.3.2. Intra-Subject Variability

The intra-subject variability refers to the differences in ECG signals from the same individual or within a single ECG signal. It is important for a biometric modality to have low intra-subject variability, as well as high inter-subject variability and stability over time. There are several sources of intra-individual variability such as chest electrode position and respiration. While the former induces variation between ECGs of the same individual, the latter induces variability within a particular ECG. Moreover, besides these two factors, intra-subject variability may also be induced by many other factors [ 8 ]:

• Physical Exercise: The duration of and the intervals between the different deflections of the heartbeats in an ECG signal vary with the heart rate. These changes are especially visible in the interval between the QRS complex and the T wave in situations of tachycardia (higher heart rates) or brachycardia (lower heart rates). Changes in the heart rate caused by physical exercise or meditation can, effectively, affect the electrocardiogram. While Lee at al. [ 69 ] showed that the ECG cycle became shorter after 10 min of physical exercise, running, and holding the breath for a certain period on different dates, Komeili et al. [ 54 ] also demonstrated that if, during feature selection, one investigates the features that are less affected by physical exercise, one can still achieve great biometric performance while exercising. Lee et al. [ 69 ] also investigated the ECG patterns of smokers, finding that there was also a minimal change in the ECG signal before and after smoking.
• Cardiac Conditions: Medical conditions of the heart can also interfere with the dynamics of the electrical pulse conduction and generate variability. In the scope of biometrics, many databases consisting of ECG signals from patients with cardiac conditions have been used. One of the most studied conditions is arrhythmia, which causes wide variations in the heart rate across time. Chiu et al. [ 82 ] mentioned that the low accuracies can be justified by unstable QRS-complexes, causing extracted features to change. However, Becerra et al. [ 78 ] stated that even though cardiac conditions affect the performance of the system, accuracies can be higher for some classifiers. Moreover, Ghazarian et al. [ 83 ] achieved different accuracies for different heart conditions, meaning that feature selection and classification optimization should be performed considering different cardiac conditions.
• Posture: Postures like standing or lying down differ widely on the position and shape of internal organs. The heart is also affected by this, and changes its position in the thorax, and thus its position in reference with the electrode placement, which causes variations in the collected ECG signal. The vast majority of the ECG acquisitions are performed with the subject lying down at rest. However, Porée et al. [ 47 ] stated that there is no requirement or advantage to compare only ECG shapes in supine resting conditions. Moreover, Raj et al. [ 16 ] presented more accurate performances while standing (which is a more realistic biometric scenario) than while sitting. Furthermore, Wahabi et al. [ 72 ] proved that the performance degrades if the training and testing signals are not from the same position, meaning that the enrollment task should be performed in different positions.
• Emotions: The sympathetic and parasympathetic systems of the autonomous nervous system work to increase or reduce the heart rate, respectively. These systems are under the direct influence of psychological states and thus, under stress, fear, and other strong emotions, fatigue, or drowsiness, the heart rate and the ECG signal can be affected. Even though some researchers still achieved good performances with different levels of stress and anxiety (91% and 97% for [ 18 , 73 ], respectively), some others proved a slightly negative impact of emotions on the performance of a biometric system. As mentioned above, Zhang et al. [ 75 ] found a reduction of the true positive rate with ECG signals acquired during high pressure and lack of sleeping situations, whereas the eating and health status did not affect the system. Thus, since there is still no consensus on whether emotions (stress, anxiety, levels of sleeping, levels of thirst and eating, etc.) negatively impact the performance of a biometric system or not, researchers should further investigate these conditions within their systems. However, we must note that it is difficult to design an experimental setup that can induce the same emotion in every subject, as different characters, varying moods, and the inability to accurately self-report an emotional experience may significantly affect the outcome of such a study.

4.3.3. Artefacts and Noise

• Electrode Material: The most used electrodes are silver/silver–chloride coated electrodes with well known frequency-band characteristics and temperature stability. However, different materials can have different characteristics affecting the recording: temperature drift can cause variation of the baseline (low frequency oscillations) while frequency-band modifications can cause attenuation of high frequencies [ 52 ].
• hlSensor Location: In the biometric scenario, it is common to use single ECG recording with only two electrodes placed in non-standard locations which can lead to different morphology of the ECG recording with respect to the standard recordings [ 73 ].
• Power-line Interference: Capacitive coupling with power lines can induce a superimposed (distorted) harmonic signal whose amplitude can obscure the morphological characteristics of ECG: notch filtering is usually used to remove this interference [ 52 ].
• Baseline Drift: Respiration causes changes in thorax volume/electrical impedance, therefore causing the isoelectric level to change slowly (in low frequencies ranges): this artefact is usually removed via low pass filtering [ 52 ].
• Movement Artefacts: The contact interface between skin/electrode is subjected to electrochemical reactions of ionic chemical species under the electrode inducing half-cell potentials of the order of 1V or less. However, patient movements can change the electrode position and the chemical concentrations of these species can vary leading to different half-cell potentials and therefore different iso-electrical (baseline) levels [ 52 ].

5. Conclusions

There are several challenges that must be addressed in order to effectively use ECG as a biometric trait. While many studies have been conducted in reviewing the most popular ECG feature sets and in highlighting similarities and differences among features and classification techniques, there is a lack of research on the data acquisition protocol [ 113 ], which is the focus of this work. Regarding the population size, the majority of the studies have been conducted on a small population (a few tens of subjects). Therefore, the applicability of ECG biometric recognition on a large scale was not yet proven. Moreover, almost all studies ignored the variability of the ECG during life span (i.e., variability induced by work, ageing, sport activity, etc.); besides, only a few studies considered the applicability of these techniques when subjects suffer from pathological conditions. ECG recognition in pathological subjects is another aspect worth additional investigation. Finally, it must be emphasized that, while guidelines are available for ECG acquisition in the clinical scenario, there is still a lack of standardization on ECG acquisition (number of leads and their positioning, sampling frequency, number of bits, filtering, type of electrodes, etc.) for biometric applications. However, ECG databases for biometric recognition should, ideally, include recordings at a given sampling frequency and condition from the same subjects in different circumstances (e.g., relaxed, during and after physical training) and for a period of several years [ 114 ].

Despite the potential of the ECG to be used as one of the main biometric traits, there are still some challenges that need to be solved, especially regarding the acquisition. Quality research is key to addressing the open issues, taking the advantage of current opportunities and proposing increasingly competitive and applicable ECG biometric systems. Furthermore, this research presents a valuable contribution to the field, emphasizing the importance of ECG data acquisition conditions, not only for biometric recognition, but also for other research topics such as disease detection.

Funding Statement

This work was funded by national funds through FCT—Fundação para a Ciência e a Tecnologia, I.P., under the Scientific Employment Stimulus—Individual Call—CEECIND/03986/2018 (R.S.) and the PhD grant UI/BD/153605/2022 (T.M.C.P.), and is also supported by the FCT through national funds, within IEETA/UA R&D unit UIDB/00127/2020, IBEB Strategic Program UIDB/00645/2020, and CICECO-Aveiro Institute of Materials, UIDB/50011/2020, UIDP/50011/2020 and LA/P/0006/2020 (PIDDAC). This work was co-funded by the European Regional Development Fund (FEDER), through Portugal 2020, under the Operational Competitiveness and Internationalization (COMPETE 2020) and Lisboa 2020 programs (grants no. 069918 LISBOA-01-0247-FEDER-069918 POCI-01-0247-FEDER-069918, “CardioLeather”), and CENTRO 2020 program CENTRO-01-0247-FEDER-113480 - “ELIPFOOTSENSE”.

Author Contributions

Author Contributions: Conceptualization, T.M.C.P., R.C.C., V.S. and R.S.; Investigation, T.M.C.P.; Formal analysis, T.M.C.P.; Methodology, T.M.C.P. and R.S.; Supervision, R.C.C., V.S. and R.S.; Writing—original draft, T.M.C.P.; Writing—review & editing, R.C.C., V.S. and R.S. All authors have read and agreed to the published version of the manuscript.

Institutional Review Board Statement

Informed consent statement, data availability statement, conflicts of interest.

The authors declare no conflict of interest.

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Biometric Recognition: Challenges and Opportunities (2010)

Chapter: 5 research opportunities and the future of biometrics, 5 research opportunities and the future of biometrics.

The first four chapters of this report explain much about biometric systems and applications and describe many of the technical, engineering, scientific, and social challenges facing the field. This chapter covers some of the unsolved fundamental problems and research opportunities related to biometric systems, without, however, suggesting that existing systems are not useful or effective. In fact, many biometric systems have been successfully deployed. For example, hand geometry systems serve to control access to, among others, university dorms, nuclear power plants, and factories, where they record time and location. 1 Automated fingerprint identification systems (AFISs) integrate automatic and manual processes in criminal justice applications and civilian applications such as national identity systems.

An emerging technology such as biometrics typically confronts unrealistic performance expectations and is sometimes unfairly compared with approaches such as passwords that are not really alternatives. An effective biometric solution does not have to be—nor can it be—100 percent accurate or secure. For example, if there exists a 1 percent possibility of successful “buddy punching” (signing in for a friend or colleague), a hand geometry system can easily be seen as preventing 99 percent of such

fraud. A particular application demands not perfection but satisfactory performance justifying the additional investments needed for the biometric system. In any given case, the system designer should understand the application well enough to achieve the target performance levels.

Nevertheless, solutions to the problem of recognizing individuals have historically been very elusive, and the effort needed to develop them has consistently been underestimated. Because humans seem to recognize familiar people easily and with great accuracy, such recognition has sometimes incorrectly been perceived as an easy task. Considering that a number of governments around the world have called for the nationwide use of biometrics in delivering crucial societal functions such as passports, there is an urgent need to act. Excepting for their application in national forensic AFISs, biometric recognition systems have never been tried at such large scales nor have they dealt with the wide use of nonforensic sensitive personal information. The current performance of some biometric systems—in particular with regard to the combination of error rate, robustness, and system security—may be inadequate for large-scale applications processing millions of users at a high throughput rate.

If there is a pressing public need for these applications, and if it is determined that biometric systems and technologies are the most appropriate way to implement them, then our understanding of the underlying science and technology must be robust enough to support the applications. 2 There is no substitute for realistic performance evaluations and sustained investment in research and development (R&D) to improve human recognition solutions and biometric systems. 3 The rest of this chapter outlines a research agenda focusing on (1) technical and engineering considerations, (2) social challenges, and (3) broader public policy considerations. The chapter concludes with a high-level overview of what constitutes a well-designed biometric system.

TECHNOLOGY AND ENGINEERING RESEARCH OPPORTUNITIES

In recent years several research agendas for biometric technologies and systems have set important challenges for the field. 4 The issues and

research opportunities raised in this chapter are meant to complement, not replace or supersede these other articulations. Indeed, the U.S. government has created or funded several interdisciplinary, academia-based research programs that provide an institutional foundation for future work. The focus of this report has been on broad systems-level considerations, particularly for large-scale applications, and the technical challenges outlined in this Chapter reflect that focus. But as these other agendas demonstrate, there are numerous opportunities for deeper understanding of these systems at almost every level. This section lays out several technical and engineering areas the committee believes would benefit from sustained research and further investigation: human factors, understanding the underlying phenomena, modality-related technical challenges, opportunities to advance testing and evaluation, statistical engineering aspects, and issues of scale.

Human Factors and Affordance

Because biometric technologies and systems are deployed for human recognition applications, understanding the subject-technology interface is paramount. A key piece of the biometric recognition process is the input of the human characteristic to be measured. With the exception of recent work at NIST-IAD 5 and at Disney, very little effort has been expended on the “affordance”—the notion that what is perceived drives the action that occurs, or, put another way, that form can drive function—of biometric systems. 6 Biometric systems should implicitly (or explicitly) suggest to the user how they are to be interacted with. 7

International standards, such as the International Organization for Standardization (ISO)/IEC 19795 series, 8 generally contain an informative annex on best practices for data collection for the modality under consideration. For example, ISO/IEC 19794-5 contains an annex on best practices for face images specifying that full frontal face poses should be used and rotation of the head should be less than +/− 5 degrees from frontal in every direction. This requirement presents an affordance challenge that has not yet been adequately addressed by the technologies—namely, how can a system be designed to suggest to the user a pose that meets this requirement? (The committee was told about one system that presented an image that, when viewed from the proper angle, was clearly visible to the user.) Similar challenges exist with every modality/application combination and will require a modality- and application-specific set of solutions.

“Quality” has been used to indicate data collected in compliance with the assumptions of the matching algorithms, such that recognition performance of the algorithm can be maximized, which means that “affordance” and data “quality” are tightly linked. System operators and administrators face their own challenges when interfacing with the systems. How should the interfaces of attended systems be designed so that the operator knows how and when to collect proper images, how to recognize when poor-quality images have been collected, and how to guide the data subject in making better presentations? Very little research in this area has been conducted, and there is opportunity for significant progress.

Distinctiveness and Stability of Underlying Phenomena

There are many open questions about the distinctiveness of the underlying biometric traits in these systems and about human distinctiveness generally. One typical assumption in the design of most biometric systems has been that characteristics, if properly collected, are sufficiently distinctive to support the application in question. This assumption has not, however, been confirmed by scientific methods for specific biometric characteristics, either by prospectively collecting and analyzing biometric samples and feature patterns or by exploiting databases of samples or feature patterns assembled for other purposes. A broad and representative sampling of the population in which distinctiveness is being evaluated should be obtained and a minimum quality specification should be set to which biometric samples should conform. More generally, the development of a scientific foundation for reliably determining the distinctive-

ness of various biometric traits under a variety of collection modes and environments is needed.

In other words, what is the effective limit on accuracy for a specific biometric trait in a realistic operating environment? This becomes a particularly important question at scale—that is, when the systems are expected to cope with large user populations and/or large reference databases. Even in DNA analysis, there has been controversy and uncertainty over how to estimate distinctiveness. 9 In biometric systems, “ground truth”—the collection of facts about biometric data subjects and recognition events to allow evaluation of system performance—is challenging, particularly for passive surveillance systems, where failures to acquire may be difficult to detect. 10

There are also open questions about the stability of the underlying traits—how persistent (stable) will a given individual’s biometric traits be over time? Some biometric traits, such as fingerprints, appear to be reasonably stable, but others, such as facial characteristics, can change significantly over even short periods of time. Depending on the capture and matching algorithms, changes in a trait over time may or may not have an effect on system performance and whether that person is appropriately recognized. Understanding more about the stability of common biometric traits will be important, especially if biometric systems are deployed for comparatively long (years or decades) periods of time.

All of this suggests several avenues of research that could strengthen the scientific underpinnings of the technology. There needs to be empirical analysis of base-level distinctiveness and the stability of common biometric modalities, both absolutely and under common conditions of capture, and research into what types of capture and what models and algorithms produce the most distinguishable and stable references for given modalities. Further, the scalability of various modalities under different capture and modeling conditions must be studied. The individuality of biometric identifiers, their long- and short-term physiological/pathological variations, and their relationship to the user population’s genetic makeup all merit attention as well.

Modality-Related Research

Every biometric system relies on one or more biometric modalities. The choice of modality is a key driver of how the system is architected, how it is presented to the user, and how match vs. nonmatch decisions are made. Understanding particular modalities and how best to use the modalities is critical to overall system effectiveness. Research into several interrelated areas will bring continued improvement:

Sensors. Reducing the cost of sensor hardware; improving the signal-to-noise ratio, the ease of use and affordability, and the repeatability of measures; and extending life expectancy.

Segmentation. Improving the reliability of identifying a region of interest when the user presents his or her biometric characteristics to the system—for example, locating the face(s) in an image or separating speech signal from ambient noise.

Invariant representation. Finding better ways to extract invariant representation (features) from the inherently varying biometric signal—that is, what kind of digital representation should be used for a face (or fingerprint or other feature) such that the trait can be recognized despite changes in pose, illumination, expression, aging, and so on.

Robust matching. Improving the performance of the matching algorithm in the presence of imperfect segmentation, noisy features, and inherent signal variance.

Reference update. Developing ways to update references so that they can account for variations and the aging of reference data in long-lived systems.

Indexing. Developing binning and partitioning schemes to speed up searches in large databases.

Robustness in the face of adversaries. Improving robustness to attacks, including the presentation of falsified biometric traits (perhaps, for example, through automated artifact detection).

Individuality. Exploring the distinctiveness of a particular biometric trait and its relationship to the matching performance. Does information about distinctiveness serve to increase understanding of the effective limits on matching performance, for example?

In addition to the general challenges described above, there are also challenges specific to particular biometric modalities and traits. While the following discussion does not describe all the challenges for each modality, it does offer some potentially fruitful avenues of investigation for the most common ones.

An ongoing challenge for facial recognition is segmentation—distinguishing facial features from surrounding information. Another signifi-

cant challenge for it is invariant representation—that is to say, finding a representation that is robust and persistent even when there are changes in pose, expression, illumination, and imaging distance, or when time has passed.

Specific challenges with respect to fingerprints include reducing the failure to enroll (FTE) and failure to acquire (FTA) rate, perhaps through the design of new sensors, artifact detection, image quality definition and enhancement, and high-resolution fingerprint matching. Fingerprint-based biometric systems could also be improved by increasing the speed of capture and minimizing contact, particularly for 10-print systems.

Iris recognition systems present R&D opportunities in the following areas: sensors; optimization of the illumination spectrum; reducing FTE and FTA rates; capturing and recognizing the iris at greater distances and with movement of the subject; and reducing the size of the hardware.

Improving speaker separation, normalizing channels, and using higher-level information (that is, beyond basic acoustic patterns) would all offer opportunities to improve voice recognition. In addition, robustness and persistence are needed in the face of language and behavioral changes and the limited number of speech samples.

Information Security Research

In many applications, biometric systems are one component of an overarching security policy and architecture. The information security community is extensive and has long experience with some of the challenges raised by biometric systems, which gives it a real opportunity for fruitful and constructive interaction with the biometrics community. Biometric systems pose two kinds of security challenges. The first is the use of biometrics to protect—provide security for—information systems. For what types of applications and in which domains is an approach incorporating biometric technologies most appropriate? This is a question for the broader information security community as well as the biometrics community and requires that we understand the goals and needs of an application to ascertain whether a biometrics-based approach is useful.

Assuming that a biometrics system is in place, the second security challenge is the security, integrity, and reliability of the system itself. 11 Information security research is needed that addresses the unique problems of biometric systems, such as preventing attacks based on the presentation of fake biometrics, the replay of previously captured biometric samples, and the concealment of biometric traits. Developing techniques

for protecting biometric reference information databases to avoid their use as a source of fake biometrics is another area for such research. Decision analysis and threat modeling are other critical areas requiring research advances that will allow employing biometric systems more fully across a range of applications.

Testing and Evaluation Research

Testing and evaluation are an important component in the design, development, and deployment of biometric systems. Several areas related to the testing and evaluation of biometric systems are likely to prove fruitful. This section describes a few of them. While there has been significant work on testing and evaluating a variety of approaches, 12 it is the committee’s view that an even broader approach has merit. Moreover, while standardized evaluations of biometric systems are highly useful for development and comparison, their results may not reliably predict field performance. Methods used successfully for the study and improvement of systems in other fields (for example, controlled observation and experimentation on operational systems guided by scientific principles and statistical design and monitoring) should be used in developing, maintaining, assessing, and improving biometric systems. (See Chapter 3 for lessons that may be applicable from other domains.) The work over the last decade within the international standards community to reach agreement on fundamental concepts, such as how error rates are to be measured, has clarified the application of test methods under the usual laboratory conditions for biometric systems deployments. 13 Guidance for potential deployers of biometric systems on what is even a useful and appropriate initial set of questions to ask before getting into the details of modalities and so forth, as developed by a number of groups, has

proven particularly useful. 14 And of course, in addition to the technical questions that need to be addressed, there are issues regarding how to measure cost over the life cycle of the system and how to assess potential and actual return on investment (ROI). Unfortunately, ROI analysis methodologies and case studies have been lacking in comparison to other types of assessments. See Box 5.1 for a brief discussion and example of an ROI assessment.

Ultimately, determining the performance of an operational system requires an operational test, because adequately modeling all of the factors that impact human and technology performance in the laboratory is extremely difficult. Although the international standards community has made progress in developing a coherent set of best practices for technology and scenario testing, guidelines for operational testing are still under development and have been slowed by the community’s general lack of experience with these evaluations and a lack of published methods and results. 15 Designing a system and corresponding tests that can cope with ongoing data collection is a significant challenge, making it difficult for a potential user of biometric systems, such as a federal agency, to determine how well a vendor’s technology might operate in its applications and to assess progress in biometric system performance. Careful process and quality control analysis—as distinct from traditional, standardized testing of biometric systems that focuses on match performance for a test data set—at all stages of the system life cycle is essential. In addition, testing methods and results should be sufficiently open to allow disinterested parties to assess the results.

Test Data Considerations

One challenge meriting attention is test data for biometric systems. Designing large-scale systems requires large test data sets that are representative of the subject population, the collection environment, and system hardware expected in the target application. How does one determine which user population will be representative of the target application? The committee believes it is unlikely that being representative of the target application is the same as being representative of the population as a whole, because the population that should be considered will vary depending upon the ultimate application for which the system is

used. Legal and privacy concerns have limited the collection and sharing of both test and operational data (for example, various data sets collected by the U.S. government) with researchers, 16 raising the question of whether biometric data can be made nonidentifiable back to its origin. 17 If it cannot, could synthetic biometric data be created and used in lieu of real biometric data? 18 If the latter is possible, does the use of synthetic (imagined) data offer any scientific validity in assessing performance of a system using real data? When test results are available, who has access to them? These and related questions merit attention from not just the T&E community but the broader biometrics communities as well.

Usability Testing

Many factors related to usability can affect system effectiveness and throughput and may also affect how well the system performs its recognition tasks. Testing and evaluation mechanisms are therefore needed that provide insight into how well a system under consideration handles a variety of user interface expectations.

Despite the recent focus of NIST’s information access division on usability testing, there is still major work to be done. One potential area of investigation is to incorporate into the design of the interface information on the expected motor control and cognitive capabilities of the user popu-

lations. Such information would allow the use of public health statistics to estimate the percentage of the general population (or subpopulation) that would be expected to have either cognitive or physical difficulties using the systems. By incorporating this understanding of the skills expected of users, designers and developers could tune the interfaces in ways that would increase their usability.

Usability is affected by other factors as well. For example, some unknown percentage of the population has a condition in which the fingers do not possess the usual friction ridges central to the functioning of fingerprint-based biometric system. In addition, some unknown (but believed to be nonzero) percentage of the population has either no irises or irises of unusual shape. When setting baseline error rates, it is important

to have estimates of the percentage of the population lacking the required trait, because this lack interacts with the design of sensors and algorithms. It may well be that each modality will have lower error rate bounds that cannot be improved upon by better sensors, algorithms, or collection procedures. More research is needed to understand this. Such questions are related to the distinctiveness and stability of the underlying biometric traits, discussed above.

Other usability considerations relate to the ease of participation. Is the system designed to take into account user needs (such as tables on which to set their items if necessary) and physical differences (such as height and weight)? What kinds of user assistance should be provided? What accommodations could be made for people who are unfamiliar with the system,

or, conversely, for people who are very familiar with it (much as toll pass transponder users can use dedicated lanes on highways)?

ROI Analysis Methodologies and Case Studies

Determining the potential ROI and identifying which system characteristics contribute is an important means of evaluating any biometrics deployment. In addition to how well a system meets its requirements, there are issues about measuring cost over the life cycle of the system and assessing potential (and actual) ROI. There are relatively few ROI analysis methodologies and case studies. The research opportunity here is to develop methods for examining likely costs and cost savings that take into account the technical life cycle as well as ongoing maintenance and usage costs.

Evaluative Frameworks for Potential Deployers

In addition to system and technology tests, there is a significant opportunity to develop an evaluative model that would guide potential procurers and users of biometric systems. Guidance for potential users of biometric systems on an appropriate initial set of questions to ask before getting into the details of modalities and so forth has proven particularly useful. 19

Testing When Data Changes

Designing a system and tests that can cope with ongoing data collection after it has been deployed is a significant challenge. The characteristics of the data may change from what was assumed during testing. This could be due to changes in the technology, changes in the user population, changes in how the system is used, or all of the above. Such potential changes to the data make it a challenge for potential users of biometrics, such as federal agencies, to readily determine how well a given vendor’s technology might operate for the agency’s applications over time. Developing testing and evaluation methodologies that can account for such potential changes or offer information on how a system’s performance

might change in the event of significant changes to the data characteristics is an opportunity for further investigation.

Operational Testing

Finally, operational testing is problematic in that most existing systems do not retain the data needed to determine error and throughput rates. Each system collects and stores different data in application-specific ways. Additionally, ground truth (all of the relevant facts about all participants) cannot be known in real applications with arbitrary user populations. Privacy rights of the data subjects may prevent using collected data for testing purposes. Lastly, because system operators may not wish others to know about operational performance for reasons of security, very few operational test results have ever been published. The ISO/IEC JTC1 SC37 standards committee has been working for a number of years on basic guidance for operational testing, but progress on this standard, to be known as ISO/IEC 19795, Part 6, has been slow, reflecting the inherent difficulties in making general statements about operational tests.

Systems-Level Statistical Engineering Research

In addition to the modality-related technical challenges outlined above, there are broad systems-level considerations to take into account. In particular, statistics and statistical engineering offer opportunities for progress and the development of design principles and model designs for operational testing of biometric systems and experimentation with modifiable system parameters. This section outlines some potential research questions in statistical engineering and biometric systems that merit attention.

Statistical approaches come into play with respect to the user populations, including cross-sectional and longitudinal studies of the variability of various biometric modalities over time, the association of biometrics with demographic and medical factors, and the effects of demographic factors and physical characteristics on failure to acquire and error rates.

Another systems-level consideration is error rates in biometric systems, including the following topics:

The relative contribution to error of different phases and components of biometric recognition (on an algorithm-by-algorithm basis, because error rates and their causes are algorithm specific);

The potential for algorithm-specific quality control measures to reduce error rates in varying populations of data subjects;

The application of known statistical methods for analysis of cor-

related data to estimation 20 of false match and false nonmatch rates for recognition tasks; and

The investigation of new statistical models for the estimation as above but from biometric databases where information on replication is incomplete—that is, replication is known only for a subset, or some information about variation in replicate measurements is available from an external source.

Other areas of potential investigation include studies of statistical efficiency and cost-benefits of different approaches to choosing, acquiring, and utilizing multimodal biometrics of various sorts. Issues to be considered would include the relative algorithmic-dependent within-class and between-class variation of sample choice such as multiple instances of 2 fingerprints versus single instances of 10 fingerprints versus single instances of 2 fingerprints combined with two-dimensional facial imaging. The aim is to develop the most informative fusion methods based on application- and algorithm-dependent multivariate distributions of biometric features.

Research on Scale

There are many ways in which scale may manifest itself in biometric systems. These include the number of sensors in the system, the number of comparisons being performed for a given unit time or a given input sample, the number of users (including administrators and operators), the geographic spread of the system, the size of the potential user population, or any combination of these factors. Addressing issues of scale in biometric systems offers numerous opportunities for research.

For instance, one question is this: How does the number of persons who have references in the enrolled database affect the speed of the system and its error rates? For some applications and associated algorithmic approaches, the size of the database might not matter if typical operation involves only a one-to-one comparison—that is, one set of submitted samples being compared to one set of enrollment records. For large-scale identification and screening systems, sequentially performing a very large number of one-to-one matches is not effective; there is a need for efficiently scaling the system to control throughput and false-match error rates as the size of the database increases.

Typical approaches to scaling include (1) using multiple hardware

units, (2) coarse pattern classification (for example, first classifying a fingerprint into classes such as arch, tented arch, whorl, left loop, and right loop), and (3) extensive use of exogenous data (such as gender, age, and geographical location) supplied by human operators. Although these approaches perform well in practice, they come at a price. Using hardware linearly proportional to the database size is expensive. Coarse pattern classification offers substantial scaling advantages even when single measures are available and even more advantage with multiple measures—for example, fingerprints from multiple fingers—but can add to the nonmatch error rates. Use of exogenous information creates a mechanism for intentionally avoiding identification—for example, dressing as the opposite sex or appearing older—if someone is trying to avoid being recognized by the system, not to mention potential privacy compromises.

Ideally, one would like to index biometric data patterns in some way similar to that used in conventional databases in order to benefit from lessons learned in other arenas. However, due to large intraclass variation in biometric data caused by variation in collection conditions and human anatomies and behaviors, it is not obvious how to ensure that samples from the same pattern fall into the same index bin. There have been very few published studies on reliably indexing biometric patterns. 21 Efficient indexing algorithms would need to be developed for each technology/modality combination. It is unlikely that any generic approach would be applicable to all biometric measures, although efforts to understand similarities and where lessons from one type of system can be applied to another are warranted.

False-match errors generally increase with the number of required comparisons in a large-scale identification or watch-list system. As most comparisons are false (for example, a submitted sample compared to the enrollment pattern of another person), increasing the size of the database increases the number of opportunities for a false match. However, in large-scale systems it is unlikely that a sample would be compared against every possible match in the database. Instead, just as with search algorithms generally, the set of items to compare against is winnowed according to certain criteria as quickly as possible to save time and memory. Because of the nonindependence of sequential comparisons using the same sample data, coupled with architectural and algorithm design choices that are aimed at finding any matches while sustaining throughput rate and limiting active memory, the relationship between the number

of false matches and database size is a poorly understood issue meriting further investigation. 22

Although a watch-list database in a screening system is much smaller than that in a large-scale identification application, the number of continuous or active comparisons may be huge. Therefore, as in large-scale applications, the throughput and error-rate issues are also critical in screening applications. Computationally, scaling of large systems for near-real-time applications involving 1 million identities is becoming feasible, as is screening the traffic for 500 recognized identities. However, designing and building a real-time identification system involving 100 million identities is beyond our understanding. More research is needed here as well.

Social Science Research Opportunities

Biometric systems require an intimate association between people and the technologies that collect and record their biological and behavioral characteristics. This is true whether the application is overt or covert, negative claim or positive claim. It is therefore incumbent on those who conceive, design, and deploy biometric systems to consider the cultural and social contexts of these systems. Unfortunately, there are few rigorous studies of these contexts. Below is a framework for developing a portfolio of future research investigations that could help biometric systems better cope and perform within their cultural and social contexts.

Cultural and social issues arise at essentially two different levels—for the individual and for society. At the level of the individual, whether they are interacting actively or passively with a biometric system (for example, the person seeking entry to a facility), the issue is the performance of a biometric system. At the societal level, the issue is the social impact of the biometric system (for example, all are affected, either directly or indirectly, by the trade, tourism, and terrorism effects of a biometric passport).

At the level of the individual, social considerations are critical in the design, deployment and functioning of biometric systems. As we have noted, system performance may well be degraded if relevant social factors are not adequately taken into consideration. For example, religious beliefs that call for adherents to cover their faces in public make facial-recognition biometrics problematic. Thus if a biometric system is to work well for a broad range of people it must take into account behaviors resulting from

such things as religion or social convention. Every biometric system has a protocol for how it is to be interacted with. The protocol may be simple or complex, uniform in application, or tailored to the individual. Obviously, however, a good protocol for a biometrics system must recognize variations in biological features. A system based on fingerprints must have ways to gracefully accommodate a person who is missing a finger or who otherwise does not have usable fingerprints.

In addition to the design issue of affordance, previously discussed, research is needed to determine effective, appropriate, and graceful protocols, processes, and devices that gain the cooperation of participants, and the protocols and devices must be acceptable to the community. In biometric systems that are essentially surveillance systems, compliance should be thought of as more than acquiescence and should extend to gracefully (perhaps without notice) promoting the types of behavior (for example, face pose and angle) that result in useful biometric measures. Full compliance represents the ideal interaction of the participant with the biometric system from the viewpoint of the system designers. Acceptability to the community refers to the endorsement, or at least the lack of active disapproval, by significant governmental and public leadership groups. In any case, community acceptability is not guaranteed. Influential parts of the community may find biometric systems overly intrusive, unfair to certain groups, or inadequately protective of the individual’s privacy.

The dimensions of individual compliance and community acceptability are discussed next. One part of the design process for particular systems or, more realistically, for a particular class of systems might be to develop data that predict how well the biometric system will perform in a target community and on factors that may make the system more acceptable to that community. Predictive aspects may just have a statistical relationship with subject compliance or community acceptability, while acceptability factors probably have a causal relationship. Developing data in these areas will provide the evidence needed to assess the relationships.

The extent and nature of participant compliance can be discovered and confirmed using either or both of two basic research strategies: field studies using ethnographic tools such as in cultural anthropology or attitude studies of using survey methods such as are common in sociology. Some of the things that might predict participant compliance include participant attitudes toward authority, their willingness to try new technology, their adherence to certain religious or cultural beliefs, and the geographic distribution of the population. Such work could, in theory, be part of the design work for biometrics systems but is generally not done, possibly because of the expense and effort involved. Research that

sheds light on these issues would provide valuable information for those designing and building biometric systems.

Factors that motivate participant compliance can be discovered by experimental studies, essentially creating laboratory environments in which the factors can be controlled. This research paradigm is common in experimental psychology, but the extent to which such controlled studies might develop data that reflect factors encountered in operational applications is only speculative. Some candidate factors include self-interest, enforcement, inducement, social pressure, conviction, habit or practice, behavior of other actors, pleasantness of the experience, and attention to cultural norms. The more common approach is to survey data subjects who have just encountered an operational system to elicit their opinions, 23 but even this approach has rarely been applied.

Aspects that predict the extent and nature of community acceptability can be discovered and confirmed using either or both of two basic research strategies: field studies of similar deployments using ethnographic tools—as indicated for participant compliance above, or focus groups that are asked to discuss how they view various characteristics of a biometric system such as are common in marketing studies. The kinds of aspects that might be predictive of community acceptability include resemblance to existing well-tolerated systems, operated under the auspices of a respected institution, or a system that meets all legal requirements

One factor in motivating community acceptability is whether community concerns—for example, fairness, privacy, and confidentiality—are addressed. Using data from research in these and related areas, it should be possible to address a variety of relevant questions, such as: Where on the scale of purely voluntary to mandatory is a particular biometric system? In largely voluntary systems does cooperation vary by subgroup such as age, sex, or race? Does habituation lead to greater cooperation? How important is it that participants believe they or others will not be harmed? What factors influence such trust? What are effective and appropriate compliance mechanisms for biometric systems? Although it is not reasonable to expect designers of a specific system to conduct such research, these questions could be addressed as part of a more general research agenda.

PUBLIC POLICY CONSIDERATIONS AND RESEARCH OPPORTUNITIES

Numerous issues come into play beyond technical and engineering considerations in government use of biometric systems. These other issues include the following:

To what extent can the need for a biometric system be satisfied by current technical capability? Balancing mandates with maturity of systems and technology is critical. Aggressive schedules can push technology development forward but not all challenges can be addressed on short notice.

Is there sufficient flexibility and time to support the risk management needed to develop and deploy a biometric system? Governments must avoid increasing risk through overly constrained integration and testing timelines and budgets. The risks include the possibility not only that the system will fail or be compromised but also the possibility that the system will be rejected by its users or be so cumbersome or inefficient that it is withdrawn from use.

Should participation in the system be mandated? Such a mandate might foster a climate of distrust or social unrest.

What is the nature of the biometrics workforce? To the extent that biometric systems and related technologies are seen as important to meet public policy goals, is there sufficient incentive to grow and maintain the needed expertise? Training and maintaining consistent biometrics workforce has been difficult. Several organizations recently announced plans to create certification programs for professionals, but consensus must be reached on what skills are required of a professional in this area. The creation of a biometrics undergraduate program at West Virginia University is a step in the right direction. (The program has a ready customer: the FBI’s Criminal Justice Information Center in Clarksburg, West Virginia.)

The sourcing of the technology is crucial to the government’s successful deployment of technological and information systems, including biometric systems. There is an inherent danger in relying on companies with manufacturing, research, or development activities centered overseas. For biometric systems, especially, the risk is the potential for U.S. biometric data to be collected by foreign governments, inviting scrutiny of U.S. information on border control systems and other critical infrastructure by persons not cleared by the U.S. government.

The social science considerations described above may have impacts on broader public policy considerations. Systematic empirical research and factual analysis would help provide an evidence case for public

policy in this area. Some key research questions that have an impact on public policy include the following:

What lessons can be learned from environmental impact statements and privacy impact assessments that might be relevant to deciding whether social impact assessments for biometric systems are useful?

Do existing or proposed biometric systems represent a serious potential for identity theft?

How have authoritarian regimes made use of human recognition methods to assert their control over individuals? In what ways might biometric systems enable these sorts of uses? How could such a risk be mitigated?

To what extent are privacy requirements, interagency control issues, and policy constraints, or the perception thereof, inhibiting the research use and sharing of existing biometric data?

What belief sets, if any, lead to an aversion to certain biometric technologies?

A reliable and effective biometric system may be perceived as providing irrefutable proof of identity of an individual, notwithstanding the many uncertainties already mentioned, raising concerns for users. Will the information regarding biometrics-based access to resources be used to track individuals in a way that infringes on privacy or anonymity? Will biometric data be used for an unintended purpose: For example, will fingerprints provided for access control be matched against fingerprints in a criminal database? Will data be used to cross-link independent records from the same person—for example, health insurance and grocery purchases? How would a user be reassured that a biometric system is being used for the intended purpose only?

Designing information systems—not only biometric systems—whose functionality is verifiable during deployment is very difficult. One solution might be a system that meticulously records recognition decisions and the people who accessed the logged decisions using a biometric-based access control system. Such a system could automatically warn users if a suspicious pattern is seen in the system administrator’s access of users’ logs. Another solution might be biometric cryptosystems—cryptographic keys based on biometric samples. Radical approaches such as total transparency attempt to solve the privacy issues in a novel way. But there are no obviously satisfactory solutions on the horizon for the privacy problem. Additional research on the relationship between biometric (bodily) information and privacy is needed.

The privacy protections required to facilitate data collection from and about biometric systems need to be clearly established. Because many of these systems are deployed to satisfy security needs, it is reasonable to

expect that performance and vulnerability data need to be protected. For best results, the data sets for such research should be very large, contain very few errors in ground truth (metadata indexing), be appropriately randomized, and represent the populations of interest to target applications. To the extent consistent with privacy and security, the results of the studies should be published in the peer-reviewed scientific literature and the biometric samples used made widely available to other researchers.

REALIZING A WELL-DESIGNED BIOMETRIC SYSTEM

Research in the areas described above is warranted more than ever as biometric systems become widespread and are used in critical applications. This report concludes by taking a step back and presenting a vision of a well-designed biometric system that should persist even as progress is made on the challenges described earlier in this chapter.

A well-designed biometric system includes more than technology. It is a complex combination of technology, public policy, law, human processes, and social consensus. In the long term, there may be new modalities that allow recognizing human characteristics and behaviors quickly and effectively with little or no interaction on the individual’s part. Human beings may turn out to possess distinctive traits that have yet to be fully explored or that cannot be suitably represented by present technology. Some of the potential sources of suitable signals currently being pursued include inductive signature and brain waves (EKG activity). Each of these potential signal sources could bring with it a new set of societal and policy issues requiring exploration.

Even with all of these uncertainties, and even with the many intriguing open questions that merit research, the committee believes that the following framework for a well-designed biometric system will apply for the foreseeable future. Progress in such research will lead to even more well-designed systems. This framework is offered as both an evaluative tool and as a development tool.

A well-designed biometric recognition system will have (at least) the following characteristics:

The system will be designed to take into account that no biometric characteristic is entirely stable and distinctive. In other words, it will take into account that biometric similarity represents a likely, not a definitive, recognition and and that the corresponding is true for a failure to find similarity. In particular, presumptions and burdens of proof will be designed conservatively, with due attention to the system’s inevitable imperfections.

The policies of such a system will recognize that any claimed probabilities of correctness depend on external assumptions about dynamic

presentation distributions, and that these assumptions, whether subjective or based on estimates from past internal or external data, are fallible. It will enable system operators and users to recognize that biometric information has a life cycle. Biometric information is collected or modified during, for example, enrollment, recognition, and so on. But policies should also recognize that changes in the biometric characteristics of the individual can lead to incorrect or failed recognition.

The system will be designed so that system operators and users recognize that some inaccurate information may be created and stored in the databases linked to biometric references, and that over time information in these databases will become out of date. In particular, the reliability of information in the database is independent of the likelihood of correct recognition. The system will be designed to handle challenges to the accuracy of database information in a fair and effective way.

Because the system’s sensors and back-end processes are not perfectly accurate, it will need to handle failures to enroll, failures to acquire samples, and other error conditions gracefully and without violating dignity, privacy, or due process rights.

Because some individuals will attempt to force the system into failure modes in order to avoid recognition, the system’s failure modes must be just as robustly designed as the primary biometrics-based process.

The system’s security, privacy, and legal goals must be explicit and publicly stated, and they must be designed to protect against a specific and enumerated set of risks. The system will specifically address the possibility that malicious individuals may be involved in the design and/or operation of the system itself.

It will recognize that biometric traits are inherently not secret and will implement processes to minimize both privacy risks and risks of misrecognition arising from this fact.

CONCLUDING REMARKS

This report lays out a broad systems view and outlines many of the subject areas with which biometrics research intersects. The committee also describes many open research problems, ranging from deep scientific questions about the nature of individuality to vexing technical and engineering challenges. It raises questions about appropriate system architecture and life-cycle design as well as questions about public policy regarding both private sector and government use of biometric systems. It notes that biometrics is an area that benefits from analyzing very large amounts of data. These and other aspects of biometrics suggest many fruitful areas and interesting problems for researchers from a range of disciplines.

Biometric recognition—the automated recognition of individuals based on their behavioral and biological characteristic—is promoted as a way to help identify terrorists, provide better control of access to physical facilities and financial accounts, and increase the efficiency of access to services and their utilization. Biometric recognition has been applied to identification of criminals, patient tracking in medical informatics, and the personalization of social services, among other things. In spite of substantial effort, however, there remain unresolved questions about the effectiveness and management of systems for biometric recognition, as well as the appropriateness and societal impact of their use. Moreover, the general public has been exposed to biometrics largely as high-technology gadgets in spy thrillers or as fear-instilling instruments of state or corporate surveillance in speculative fiction.

Now, as biometric technologies appear poised for broader use, increased concerns about national security and the tracking of individuals as they cross borders have caused passports, visas, and border-crossing records to be linked to biometric data. A focus on fighting insurgencies and terrorism has led to the military deployment of biometric tools to enable recognition of individuals as friend or foe. Commercially, finger-imaging sensors, whose cost and physical size have been reduced, now appear on many laptop personal computers, handheld devices, mobile phones, and other consumer devices.

Biometric Recognition: Challenges and Opportunities addresses the issues surrounding broader implementation of this technology, making two main points: first, biometric recognition systems are incredibly complex, and need to be addressed as such. Second, biometric recognition is an inherently probabilistic endeavor. Consequently, even when the technology and the system in which it is embedded are behaving as designed, there is inevitable uncertainty and risk of error. This book elaborates on these themes in detail to provide policy makers, developers, and researchers a comprehensive assessment of biometric recognition that examines current capabilities, future possibilities, and the role of government in technology and system development.

READ FREE ONLINE

Welcome to OpenBook!

You're looking at OpenBook, NAP.edu's online reading room since 1999. Based on feedback from you, our users, we've made some improvements that make it easier than ever to read thousands of publications on our website.

Do you want to take a quick tour of the OpenBook's features?

Show this book's table of contents , where you can jump to any chapter by name.

...or use these buttons to go back to the previous chapter or skip to the next one.

Jump up to the previous page or down to the next one. Also, you can type in a page number and press Enter to go directly to that page in the book.

Switch between the Original Pages , where you can read the report as it appeared in print, and Text Pages for the web version, where you can highlight and search the text.

To search the entire text of this book, type in your search term here and press Enter .

Share a link to this book page on your preferred social network or via email.

View our suggested citation for this chapter.

Ready to take your reading offline? Click here to buy this book in print or download it as a free PDF, if available.

Get Email Updates

Do you enjoy reading reports from the Academies online for free ? Sign up for email notifications and we'll let you know about new publications in your areas of interest when they're released.

• Best biometrics use cases become clearer as ecosystems mature

Biometrics are for digital identity, socio-economic development, air travel and remote identity verification, but not public surveillance, the most-read news of the week on Biometric Update indicates. ID4Africa’s 10th anniversary event demonstrated the considerable progress made by many nations across the continent on providing legal and digital ID, as many graduate to establishing DPI that makes digital identity more useful to people. NADRA’s participation was a little different this year, with a new role as a biometric equipment vendor. The market for digital infrastructure in developing countries remains a controversial topic, as Huawei shows. The EU is moving ahead with the AI Act and digital wallets. SITA’s acquisition of Materna makes a big player even bigger, while Footprint’s funding round sets it up to compete for growth in the KYC and identity verification market.

Top biometrics news of the week

ID4Africa held another successful summit this week in Cape Town, South Africa, focusing on digital public infrastructure amid a range of pressing issues for the continent in legal and digital identity. One of many highlights was the announcement that Ethiopia will host ID4Africa 2025 . Addis Ababa is the home base of the African Union and one of the continent’s main travel hubs, and its turn hosting the world’s preeminent digital ID event is hotly anticipated.

NADRA introduced its domestically-developed all-in-one biometric enrollment kit and iris scanner at ID4Africa 2024 . The kit supports face, fingerprint and iris biometrics, encrypting and transmitting the data, and can also be used for identity verification, a NADRA official tells Biometric Update . A rumored expansion of CNIC issuance is not happening, the Pakistan ID agency says.

Huawei is carrying out contracts to build digital infrastructure around the world , from Kenya and Zambia to Malaysia and Thailand. Some observers are concerned about the company’s connection to China’s political elite, seeing a kind of digital Belt and Road initiative.

The AI Act has received final approval from the European Union Council, setting up an implementation path that will take three full years to complete. In that time, a new AI Office will be established. Real-time remote biometric identification technologies, such as live facial recognition, can only be used in public spaces in extreme conditions.

The EU’s Digital Identity Regulation also entered into force, building on eIDAS and setting up the issuance of national digital identity wallets to support a range of interactions. That leaves harmonizing regulation, establishing a common technical architecture, reference framework and standards on the agenda.

Italy began unveiling its national digital ID wallet just before EUDI became official, and ahead of a planned January 2025 rollout. The IT Wallet is intended to give Italians more control over their identity and personal data, while easing a wide range of online and in-person interactions.

SITA acquired Materna IPS , one of the leading global providers of hardware and software for airport biometric touchpoints. The deal expands SITA’s portfolio and gives it more options to bundle with its biometrics platform, as it pursues a strategy of offering comprehensive curb-to-gate passenger processing solutions.

Footprint has raised $13 million to expand its identity verification and KYC services , including with a new fraud prevention suite. Zoloz, Nametag, Nivo and Prembly each revealed fraud prevention deals or licenses, and Veriff and Shufti Pro are targeting copyright fraud in the music industry.

The cost of fraud perpetrated with generative AI and deepfakes is hitting the financial services industry hard . Officials in Hong Kong is warning the public about a deepfake endorsement of a dodgy crypto platform, after Sumsub warned last year that it is one of several APAC markets with spiking AI fraud. An NCC co-founder says banks are not ready to stop the threat.

Two research papers explain advances in the generation of synthetic data to train facial recognition algorithms with. A paper from Idiap researchers have developed a method for improving the diversity of synthetic datasets, while a team from Hochschule Darmstadt tackles the creation of datasets to train algorithms to recognize children over time.

Please tell us about any thought leadership pieces, webinars or other content we should share with the people who work in biometrics and the digital identity community either in the comments below or through social media.

Related Posts

Article Topics

biometric authentication  |  biometric identification  |  biometrics  |  digital ID  |  week in review

Latest Biometrics News

Lawmakers paint grim picture of us data privacy in defending apra.

The U.S. Subcommittee on Innovation, Data, and Commerce is marking up proposed legislation – and taking names. In remarks opening…

Decentralized ID presents opportunity to tackle identity verification challenges

Almost all organizations (97 percent) are currently encountering challenges with identity verification, according to a survey conducted by Ping Identity….

UN advocate lauds financial inclusion in the Philippines through digital ID

Financial inclusion is a critical objective for nations seeking to enhance economic development and resilience, particularly in regions where large…

Sierra Leone’s digital government drive gets push with data sharing MoU

Internal efforts within the government of Sierra Leone to add traction to the country’s digital government push have got a…

Canadian cops contract Idemia facial recognition

York Regional Police and Peel Regional Police have announced they are joining forces to contract and implement a facial recognition…

Digi Yatra keeps expanding, learns from growing pains

Digi Yatra Foundation CEO Suresh Khadakbhavi is on the PR trail, as India’s biometric app to facilitate quicker boarding prepares…

Leave a Reply Cancel reply

This site uses Akismet to reduce spam. Learn how your comment data is processed .

Continue Reading

Most Read This Week

• EU watchdog rules airport biometrics must be passenger-controlled to comply with GDPR
• Ethiopian capital Addis Ababa unveiled as host of 2025 ID4Africa AGM
• Real-time remote biometrics banned in EU with final green light for AI Act
• Interpol issues call for biometrics firm to provide mobile data collection devices
• Facewatch, Met police face lawsuits after facial recognition misidentification
• Thales-led consortium wins Mauritius’ digital ID card and wallet contract
• UK Biometrics and Surveillance Camera Commissioner role survives as DPDI fails
• New international standard covers forensic biometrics in lab and field environments
• Choose strategy for digital ID wallet integration, governance frameworks carefully
• Daily Biometrics News

Featured Company

• More Biometrics Companies

Biometrics Insight, Opinion

• Who is looking out for your data? Security in an era of wide-spread breaches
• ITL’s Alerts App expands biometric portfolio to integrated venue management
• Demystifying cross-border data transfer compliance for Indian enterprises
• Leveraging identity: The first line of cyber defense strategy
• Deepfake fears motivate consumers to change behavior, but misconceptions linger
• Biometrics Industry Interviews

Digital ID In-Depth

• UNHCR salutes African govts’ pledge to accelerate digital ID action for forcibly displaced
• DPI critical to effective digital transformation in government
• Digital transformation, easing access to services and NIST in the spotlight at ID4Africa Day 2
• Africa’s digital ID experiences yield lessons on engaging stakeholders, open standards
• Digital ID Biometrics

Biometrics White Papers

• Protect your business and customers from threat of account fraud
• A buyer’s guide to online identity verification
• Generative AI: How to protect your business in the golden age of fraud and disinformation
• Metrics that matter – How to evaluate identity verification technology
• Digital travel: definition and benefits
• More White Papers

Biometrics Events

• 14th Africa Bank 4.0 Summit – Pan Africa Nairobi, Kenya: May 29 - May 30, 2024
• GITEX Africa Marrakech, Morocco: May 29 - May 31, 2024
• ENBANTEC Cyber Security Conference and Exhibition Istanbul, Turkey: Jun 4, 2024
• AI & Big Data Expo North America Santa Clara, CA: Jun 5 - Jun 6, 2024
• Cyber Security & Cloud Congress North America Santa Clara, CA: Jun 5 - Jun 6, 2024
• More Biometrics Events