We describe a speechreading (lipreading) system purely based on visual features extracted from grey level image sequences of the speakers lips. Active shape models are used to track the lip contours while visual speech information is extracted from the sha ...
1996