4.1. Introduction

Facial animation for visual speech is a topic situated at the confluence of two rapidly emerging technology domains in the area of multimedia: speech technologies and 3D animation. On the one hand, speech is a visible process. In fact, even though the acoustic signal is generally sufficient to guarantee the full comprehension of a spoken message, the visual information carried by the face of an interlocutor enables some compensation of the acoustic degradations that may occur when listening conditions become adverse. This possibility of using lip reading is well known to hearing impaired people, but it is also used by regular listeners. As a consequence, some research areas interested in speech perception, recognition and synthesis have naturally focused on this bimodality. Thus, the design of anthropomorphic visual models of speech articulators enables the testing of scientific hypothesis concerning the integration of acoustic and visual information by human beings and the improvement of speech intelligibility for the current speech synthesis systems. On the other hand, given the level of quality reached today in the rendering and the anthropomorphic character of some 3D face models, the animators and the general public are likely to become less and less tolerant of inconsistencies that may subsist between facial gestures and the speech produced by these virtual actors. It therefore becomes necessary to develop animation techniques and models which account for the complexity and the specificity of speech production.

After a brief description of the current applications in...