This paper describes a speaker recognition system that uses both acoustic speech and visual speech (motion of visible articulators). Integration of acoustic and visual speech aims at improving recognition performance with regard to recognition accuracy, robustness against variability of input data, and protection against impersonation. As an initial step towards this goal, voice has been used together with still face images; this combination of vocal and facial information has resulted in better recognition accuracy than from either of the two constituents individually.
Bibliographic reference. Chibelushi, C. C. / Mason, J. S. / Deravi, R. (1993): "Integration of acoustic and visual speech for speaker recognition", In EUROSPEECH'93, 157-160.