RSS

Realistic simulation of visual speech

Investigator: James Edge Supervisor: Martin Cooke and Alan Watt (Graphics research group).

Recently much research has been conducted into the simulation of communicating virtual agents. These agents have been put to many uses varying from entertainment to human-computer interaction and speech therapy. However, none of the current systems can claim to have fully solved the complex task of visual speech simulation. Important features of speech production, such as coarticulation, are often ignored or crudely approximated. This research involves the integration of advances in facial representation and control with knowledge of speech production to produce realistic simulations of visual speech. Currently work is being conducted into the use of muscle level facial control parameters within a text-to-visual-speech(TTVS) system.