A Hierarchical Fusion Strategy In Multibiometric Authentication Systems

Résumé: Multi-biometric authentication systems that fuse information from multiple biometric sources, have gain more space, in the field of security and more precisely in the field of recognition and verification of person identities, this, due to their ability to overcome the limitations of uni-biometrics as the non-universality of the biometric traits, the noise at biometric sensors level and the large intra-user variation ... etc. In this thesis, the case of the fusion of biometric data is inspected in all these circumstances in order to release a multi-biometric system based on biometric fusion of face, fingerprint, voice, online signature or finger vein. The majority of multi- biometric systems proposed in the state of the art of authentication systems are based on the fusion or integration in serial mode or parallel mode, however, this thesis consider to explore hierarchical fusion strategy to benefit from the advantages of both of serial and parallel modes and to improve the overall recognition rate of the authentication system. In addition, biometric performance enhancement is a chalange. In this thesis, a multimodal biometric system based on hierarchical strategy of fusion, is presented. This strategy combines several biometric traits based on multi-level biometric fusion hierarchy. The multi-level biometric fusion includes a prior-to-matching fusion with optimal feature selection and an after-matching fusion based on the similarity of minimum of distances. The proposed solution enhances the overall recognition performances based on feature selection and reduction using principal component analysis (PCA) or Linear discriminant analysis (LDA).

Publié dans la revue: