Multiresolution Analysis and Geometric Measures for Biometric Identification Systems

Abstract

The authors describe here their work in two different biometric techniques: iris pattern and hand geometry. These two techniques require very different processing algorithms, achieving good results for two different security environments. While iris pattern is a high cost and very high reliability technique, hand geometry is low cost, highly accepted and provides medium/high security. Adaptive techniques have been studied for increasing hand geometry reliability while new processing algorithms, such as symmetric real Gabor filters, have been used to decrease the computational cost involved in iris pattern recognition. The target of adapting these biometric techniques for small embedded systems, comes with the design of new Biometric Systems, where the user’s template is stored in a portable storage media. This media can also be used to store some user’s sensible information, such as his health record, and new access conditions are designed to avoid the reading of this data unless the biometric verification has been performed. The media chosen has been a smart card, and a new requirement is established: biometric verification (hand or iris) should be performed inside the smart card, in order to not be able to extract the user’s template from the card.