Audio-visual interaction (3)
Speech-driven talking heads
- Wireframe models are used, such as the CANDIDE model
developed at Linkoping University, Sweden.
- To synthesize various facial expressions, the facial action coding
systems is used to generate the required trajectories of the
- Computer graphics techniques, such as texture mapping are
employed to add photorealism.
- HMMs or vector quantization techniques are used to find
speech-to-viseme mappings and to drive the talking heads.