Animation Control

Animating every subtle facial action by hand is an extremely tedious task and requires the skill of a good animator. Automatic lip synchronization is included into animation systems by a hierarchical approach [66] or by adding speech parameters [60]. A correspondence between each speech unit and a basic lip shape is established. Some coarticulation effects are integrated [28][110].

Three philosophical approaches have been taken to produce automatic facial animation: rules, analysis, and performance.

• Rule-based approach: Based on linguistics and psychological studies, a rule-based approach proposes automatic generation of facial animation. A set of rules describes the link between the intonation, emotion and facial expressions [110]. Multi-layer structures allows one to characterize various synchronous effects (lip movements, conversational signals, emotions and so on) [66].

• Analysis-based approach: This technique consists of extracting information from a live-video sequence and giving it as input to the animation system [43][134]. Such information corresponds to muscle contractions or determination of FACS Action Units. Automatic extraction of facial parameters are difficult due to the subtlety and complexity of facial deformations and muscle correlations, but is a subject of much study in the computer vision community as reported by the Planning Workshop on Facial Expression Understanding.

• Performance-based approach: By tracking various points on a live actor's face and texture-mapping images onto an underlying polygon model, real-time facial animation synthesis can be achieved with little computational cost and no analysis [151] [[[plus recent system by SimGraphics]]].