Fusion of face and voice using the Dempster-Shafer Theory for person verification

Abstract

In this paper, an algorithm for person verification is proposed. Dempster-Shafer Theory is used for face and voice fusion at the score level in order to overcome the limitations of unimodal biometric systems. Our experiments on the publicly available scores of the XM2VTS Benchmark database show a consistent improvement in performance compared to each individual modality. We have compared our approach with the sum rule and the likelihood ratio based fusion.