Abstract
The work of Bernstein and Benoît has confirmed that it is advantageous to use multiple senses, for example to employ both audio and visual modalities, in speech perception. As a consequence, looking at the speaker's face can be useful to better hear a speech signal in a noisy environment and to extract it from competing sources, as originally identified by Cherry, who posed the so-called "Cocktail Party" problem. To exploit the intrinsic coherence between audition and vision within a machine, the method of blind source separation (BSS) is particularly attractive. © 2010 Springer-Verlag.