A Facial Animation System for Generating Complex Expressions

This paper presents a novel expressive facial animation system based on a motion captured data stored in a database. Unlike common data-driven facial animation systems which only use mo-cap data of one expression at a time, our system can use data of more than one expression. Thus, the system can create a complex expression when the virtual human is talking. Using our method, the resulting facial animation can be a combination of the expressions from different faces in databases. In this way, complex expressive animation can be generated, even some extreme ones that are difficult for an actor/actress, e.g., a happy expression on the left side of the face that smoothly transits to the sad on the right. This paper will describe the system and show some experimental results.