Sécurisation des transactions et des déplacements des personnes par la biométrie multi-modale

Abstract

Multibiometric systems that fuse information from different sources are able to alleviate limitations of the unimodal biometric systems. In this thesis, we propose a multibiometric framework to identify people using their left and right wrist vein patterns. The framework uses a fast and robust preprocessing and feature extraction method. A generic score level fusion approach is proposed to integrate the scores from left and right wrist vein patterns using Dubois and Parad triangular norm (t-norm). Experiments on the publicly available PUT wrist vein dataset show that the proposed multibiometric framework outperforms the unimodal systems, their fusion using other t-norms techniques, and existing wrist vein recognition methods. Also, a new framework for score level fusion based on symmetric sums (s-sums) has been presented. These s-sums are generated via triangular norms. The proposed framework has been tested on two publicly available benchmark databases. In particular, the authors used two partitions of NIST-BSSR1, i.e. NIST-multimodal database and NIST-fingerprint database. The experimental results show that the proposed method outperforms the existing approaches for the NIST-multimodal database and NIST-fingerprint database