Speech driven facial animation using a hidden Markov coarticulation model

We present a hierarchical image based facial model which is driven from speech. It incorporates a novel modelling and synthesis algorithm for learning and producing coarticulated mouth animation. We demonstrate using the hierarchical model how animation of the entire face may be created solely from animations of the mouth and how colour may be reincorporated and reproduced compactly without explicitly being modelled. We then consider and evaluate methods of merging animations from several different facial areas for delivery of output.