Amitava Chatterjee, professor at Jadavpur University of Calcuta-India

LISSI's Meeting room, Vitry-sur-Seine

Sparse And Collaborative Representation for Solving Biometric Problems

Abstract:

In recent times, development of palm dorsa subcutaneous vein pattern (PDSVP) based biometric authentication/verification systems has evolved as a popular, comparatively new biometric solution methodology. PDSVPs lie beneath the human skin, they cannot be altered until some surgical intervention is carried out and they are believed to be quite unique that remain invariant over a long period of time. Such PDSVPs require specialized thermal imaging for image acquisition purpose. Thermal imaging of palm dorsa vein patterns is also emerging as a very attractive biometric alternative because these are least vulnerable to sensor level spoof attacks and are quite difficult to forge.
In this talk, at first, the basics of far infrared (FIR) and near infrared (NIR) imaging of palm dorsa vein patterns are covered in detail. The next part of the talk focuses on introduction of two recently proposed classes of pattern recognition algorithms, called Sparse Representation based Classification (SRC) and Collaborative Representation based Classification (CRC). In recent years, there has been an enormous thrust in research efforts to solve several classes of signal processing and image processing problems, using the concept of sparse coding or sparse representation. The talk gives a detail account of how infrared imaging of PDSVP can be successfully integrated with SRC and CRC based pattern recognition algorithms to develop real biometric system solutions. The talk concludes with a discussion about how a more critical scenario in the domain of biometrics, called the small sample size (SSS) problem, can be solved by suitably adapting SRC and CRC algorithms for PDSVP based biometrics.