Acoustic-labial Speaker Verification

This paper describes a multimodal approach for speaker ver-iication. The system consists of two classiers, one using visual features and the other using acoustic features. A lip tracker is used to extract visual information from the speaking face which provides shape and intensity features. We describe an approach for normalizing and mapping dierent modalities onto a common conndence interval. We also describe a novel method for integrating the scores of multiple classiers. Veri-cation experiments are reported for the individual modalities and for the combined classier. The performance of the integrated system out-performed each subsystem and reduced the false acceptance rate of the acoustic subsystem from 2.3% to 0.5%.