GMM-based multimodal biometric verification

In this work, we describe how biometric data can be used for person identification and verification. We rely on three categories of traits, that is speech, signature, and face. These distinguishing features or characteristics of a person, on their own, do not provide satisfactory results using well-known techniques. This is the case especially when the number of enrolled persons is large. For this reason, we develop techniques for making good use of all the three traits. In particular, we choose to follow late fusion of the scores of each single trait. The results of these techniques are quite better than using only one trait. Another goal of this work is the creation of a high quality multilingual database with video, audio, and signatures from forty seven persons.