SESAM: A biometric person identification system using sensor fusion

Abstract

In the present paper we describe the person authentification system SESAM. Person identification and verification still is a very difficult task. Using one biometric feature, i.e. the photograph or the sound of the voice, leads to goods results, but there is no reliable way to verify the classification. In order to reach robust identification and verification we are combining three different biometric cues. These cues are dynamic, i.e. the sound of the voice and the lip motion, and static, i.e. the fixed image of the face. Each branch is preprocessed and classified separately and the results are combined, e.g. in a 2-from-3 manner. The recognition of persons may be used for pure identification or can be varied to a verification system. For both cases we have done a field test to show that this approach leads to a reliable person authentification system.