Icassp2003.com

Prof. Vijayakumar Bhagavatula and Dr. Marios Savvides, Carnegie Mellon University

Abstract

Identity recognition is important in many applications including access control and surveillance. Current approaches for identity recognition are based on passwords or PINs that can be lost, stolen or forgotten. Use of biometrics (e.g., face image, fingerprints, and iris image, etc.) for identity recognition is more attractive since biometrics are integral to a person. Recently, new correlation filter-based biometric recognition algorithms have been developed that work in the spatial frequency domain (i.e., the 2-D discrete Fourier transform of the images) and offer benefits such as shift-invariance, graceful degradation and closed-form expression. These frequency domain-based methods have performed very well in the face recognition grand challenge (FRGC) and the iris challenge evaluation (ICE). The goal of this tutorial is to present the underlying theory as well as the application of correlation filters for biometric recognition. In particular, three biometric modalities (namely face image, iris image and fingerprint) will be considered, although these techniques can be easily extended to other image-type biometric modalities. This course will review the needed basic concepts from linear algebra, signal processing, optimization and pattern recognition theory to help the attendees understand the advantages and limitations of correlation filters. This course will present the design of correlation filters and summarize the performance of the correlation filter-based approaches with publicly-available biometric data bases as well as in challenges such as the face recognition grand challenge (FRGC) and iris challenge evaluation (ICE).

• Motivation, terminology and notation for biometric recognition,
• Biometric modalities, general challenges in biometric recognition,
• Basic correlation filters
• Advanced correlation filters
• Image-domain face recognition (Eigenfaces, Fisherfaces),
• Correlation filters for face recognition,
• Image-domain iris recognition,
• Correlation filters for iris recognition,
• Minutiae-based fingerprint recognition,
• Correlation filters for fingerprint recognition
• Face verification results from the face recognition grand challenge (FRGC)
• Iris verification results from the iris challenge evaluation (ICE), and
• Related issues (cancellability of biometric templates, privacy, and complexity).

Targeted Audience

Students, engineers, scientists and technology managers who want to understand how signal processing methods can be used for biometrics applications. Undergraduate-level background in probability theory, vectors and matrices, and basic signals and systems theory is assumed.

Speaker Biographies

Vijayakumar (‘Kumar’) Bhagavatula received his B.Tech., and M.Tech., degrees in Electrical Engineering from Indian Institute of Technology, Kanpur and his Ph.D. in Electrical Engineering from Carnegie Mellon University (CMU), Pittsburgh. Since 1982, he has been a faculty member in the Department of Electrical and Computer Engineering (ECE) at CMU where he is now a Professor. Professor Kumar also served as the Associate Department Head of the ECE Department from 1994 to 1996 and as the Acting Department Head from 2004 to 2005. Professor Kumar’s research interests include Automatic Target Recognition Algorithms, Biometric Recognition Methods and Coding and Signal Processing for Data Storage Systems. He has authored or co-authored one book (Correlation Pattern Recognition, Cambridge University Press, UK, November 2005), eight book chapters and more than 400 papers in these areas. He served as a Pattern Recognition Topical Editor for the journal Applied Optics. He is currently an Associate editor for IEEE Transactions on Information Forensics and Security. Professor Kumar has served on many biometrics conference program committees including the SPIE conference on Biometric Technology for Human Identification and the International conference on Biometrics. Professor Kumar is a senior member of IEEE, a Fellow of SPIE - The International Society of Optical Engineering, a Fellow of Optical Society of America (OSA) and a Fellow of the International Association of pattern recognition (IAPR).

Dr. Marios Savvides obtained his PhD in Electrical and Computer Engineering, Carnegie Mellon University in 2004. He is currently a Research Assistant Professor there with a joint appointment in Carnegie Mellon’s CyLab. His research interests are in Biometric Recognition of Face, Iris and Fingerprint and Palmprint modalities. He is a member of IEEE and SPIE and serves on the Program Committee of several Biometric conferences. He is also listed in 2006 Edition of Marquis’ Who’s Who in America. He has authored and co-authored about 70 conference and journal articles in the Biometrics area and has filed two patent applications in this field.