Biometric Identification from Facial Sketches of Poor Reliability: Comparison of Human and Machine Performance

Facial sketch recognition refers to the establishment of a link between a drew representation of a human face and an identity, based on information given by a eyewitness of some illegal act. It is a topic of growing interest, and various software frameworks to synthesize sketches are available nowadays. When compared to the traditional hand-made sketches, such sketches resemble more closely the appearance of real mugshots, and led to the possibility of using automated face recognition methods in the identification task. However, there are often deficiencies of witnesses in describing the subjects’ appearance, which might bias the main features of sketches with respect to the corresponding identity. This chapter compares the human and machine performance in the task of sketch identification (rank-1 identification). One hundred subjects were considered as gallery data, and five images from each stored in a database. Also, one hundred sketches were drew by non-professionals and used as probe data, each of these resembling an identity in the gallery set. Next, a set of volunteers was asked to identify each sketch, and their answers compared to the rank-1 identification responses given by automated face recognition techniques. Three appearance-based face recognition algorithms were used: (1) Gabor-based description, with \(\ell _2\) norm distance; (2) sparse representation for classification; and (3) eigenfaces. The sparse representation for classification algorithm yielded the best results, whereas the responses given by the Gabor-based description algorithm were the most correlated to human responses.

[1]  D. Donoho,et al.  Atomic Decomposition by Basis Pursuit , 2001 .

[2]  Debotosh Bhattacharjee,et al.  Geometric feature based face-sketch recognition , 2012, International Conference on Pattern Recognition, Informatics and Medical Engineering (PRIME-2012).

[3]  S. Clark,et al.  Stress, interviewer support, and children's eyewitness identification accuracy. , 2014, Child development.

[4]  Hanqing Lu,et al.  A nonlinear approach for face sketch synthesis and recognition , 2005, 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'05).

[5]  Anil K. Jain,et al.  Matching Composite Sketches to Face Photos: A Component-Based Approach , 2013, IEEE Transactions on Information Forensics and Security.

[6]  Richard Wiseman,et al.  The gestural misinformation effect: skewing eyewitness testimony through gesture. , 2013, The American journal of psychology.

[7]  Rama Chellappa,et al.  Secure and Robust Iris Recognition Using Random Projections and Sparse Representations , 2011, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[8]  V. Bruce,et al.  A comparison of two computer-based face identification systems with human perceptions of faces , 1998, Vision Research.

[9]  Barbara Tversky,et al.  Biased Retellings of Events Yield Biased Memories , 2000, Cognitive Psychology.

[10]  J. Daugman.  Two-dimensional spectral analysis of cortical receptive field profiles , 1980, Vision Research.

[11]  B. V. K. Vijaya Kumar,et al.  Illumination Tolerant Face Recognition Using a Novel Face From Sketch Synthesis Approach and Advanced Correlation Filters , 2006, 2006 IEEE International Conference on Acoustics Speech and Signal Processing Proceedings.

[12]  K. Vinay,et al.  Face Recognition Using Gabor Wavelets , 2006, 2006 Fortieth Asilomar Conference on Signals, Systems and Computers.

[13]  A. Memon,et al.  Are Two Interviews Better Than One? Eyewitness Memory across Repeated Cognitive Interviews , 2013, PloS one.

[14]  D. Donoho.  For most large underdetermined systems of linear equations the minimal 𝓁1‐norm solution is also the sparsest solution , 2006 .

[15]  Terence Sim,et al.  Do you see what i see? A more realistic eyewitness sketch recognition , 2011, 2011 International Joint Conference on Biometrics (IJCB).

[16]  John G. Daugman,et al.  Complete discrete 2-D Gabor transforms by neural networks for image analysis and compression , 1988, IEEE Trans. Acoust. Speech Signal Process..

[17]  Anil K. Jain,et al.  Matching Forensic Sketches and Mug Shots to Apprehend Criminals , 2011, Computer.

[18]  LinLin Shen,et al.  A review on Gabor wavelets for face recognition , 2006, Pattern Analysis and Applications.

[19]  M. Turk,et al.  Eigenfaces for Recognition , 1991, Journal of Cognitive Neuroscience.

[20]  Xiaogang Wang,et al.  Face Photo-Sketch Synthesis and Recognition , 2009, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[21]  Allen Y. Yang,et al.  Robust Face Recognition via Sparse Representation , 2009, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[22]  Dennis F. Dunn,et al.  Optimal Gabor filters for texture segmentation , 1995, IEEE Trans. Image Process..

[23]  R. Bull,et al.  Archival analyses of eyewitness identification test outcomes: what can they tell us about eyewitness memory? , 2014, Law and human behavior.

[24]  Garrett L. Berman,et al.  Adult Eyewitness Testimony: Conceptual, practical, and empirical issues associated with eyewitness identification test media , 1994 .

[25]  Joni-Kristian Kämäräinen,et al.  Invariance properties of Gabor filter-based features-overview and applications , 2006, IEEE Transactions on Image Processing.