| Computer facial animation is a new topic in virtual reailty field, the main research contents including displaying the reality facial expression and generating animation sequence in real time. Based on the standard of facial animation in MPEG-4, a new method of realizing 3D facial animation is proposed in this thesis. This method is not only capable of displaying various reality facial expressions but also capable of being driven by plain text to generating animation sequence in real time.Considering different regions of face have different modes of motion, many technologies are adopted in this thesis. Use piecewise linear function to simulate the curvilinear motion of the skin. Calculate the distance between vertices with geodesic and smooth the surfaces with Radial Basis Function. Rigidly move the meshes of eyes, for those meshes rotate only. And with the result from physiotogy, simulate the motion of gazing. Add the muscle vector to the mesh to constrain the feature points'movement with the model of linear muscle and sphincter.This thesis uses the third part Text to Speech system to convert text to speech, and acquire the data about time and viseme during the procedure. Then generate visual speech animation in real time with the library of viseme. In order to improve the smooth and reality of transmition, the co-articualtion is discussed.Finally, a facial animation system is developed to check the method proposed before. |