Système De Reconnaissance Biométrique Multimodal

Résumé: Biometrics refers to the automatic identification of a person based on physiological or behavioral characteristics, such as fingerprint, face, voice, gait, etc. However, the single mode biometric system suffers from several limitations, such as lack of completeness and susceptibility to spoofing attacks. To mitigate these problems, information from different biometric sources is combined and these systems are known as multimodal biometric systems. In this thesis, we propose to optimize Cuckoo Search (CS) and Genetic Algorithm (GA) as two evolutionary approaches to combine facial and voice modalities at the conformation result level. The effectiveness of these two methods was compared to that obtained using PSO and SVM in previous results from other subjects. The well-known Min-Max normalization technique is used to convert individual match scores to a common range before fusion occurs. The proposed schemes are empirically evaluated on publicly available score datasets (XM2VTS and TIMIT) and provided that the data is of clean quality. It is evident in this study that by deploying these fusion techniques, the validation error rates (EER) can be significantly reduced.

Publié dans la revue: