Using local features based face experts in multimodal biometrics identification systems

Abstract

Using local features generally provides higher accuracies compared to a global feature vector in face identification. In this study, taking into account the fact that better multimodal systems generally include individually good experts, multimodal identification using speech and local feature based face experts is studied. Both spPCA and mPCA are considered for this purpose. Experiments on XM2VTS and BANCA databases have shown that the local features based face experts not only provide better individual accuracies but also boost the performance of the multimodal identification system. ©2009 IEEE.