The Research Of Facial Expression Animation Based On Motion Capture Data

Facial animation aims at producing expressive and plausible animations of a 3D facial model. By now, there are many 3D facial acquisition techniques available. Although motion capture is able to capture subtle facial movement from a performer, it is always a challenging task to implement realistic facial animation.As a special kind of facial expression animation, facial expression cloning is mainly that the motion vector data of the existing vertices on the original model is relocated to the target model which may have different topology structure compared with the original model, and at the same time, keeping the relative motion of the original facial animation and dynamic characteristics. In recent years, realistic facial expression cloning technology is one of hot spots in the field of computer, and widely used in games, film production and other industries.Firstly, based upon motion capture, a semi-automatic generation technique for fast individual facial animation is presented. While capturing the facial expressions from the performer, a camera can be used to record her/his front face as a texture map. Then, the Radial Basis Function(RBF) technique is utilized to deform a generic facial model and the texture is remapped to generate the personalized face. Finally, partitioning the individual head into three regions, the RBF is implemented to deform each region respectively based on the captured facial expression data, and the three parts are combined into one whole to construct the final face model after shape blending. Our results show that the technique is efficient to fast generate realistic facial animation.Secondly, in view of the reality of facial expression cloning and efficiency of expression reconstruction, a novel method based on motion capture data is proposed. After capturing the data of six fundamental expressions, we normalize these data to make them in the same range. Then 41 points are chosen in critical areas of facial expression and we finally get cloning expression using Laplace deformation algorithm with convex weight which can preserve the details of facial expression to avoid the low fidelity of uniform weights and unstable calculation of cotangent weights. Experimental results show that our method can generate realistic and natural expression animations and the efficiency of facial expression cloning is improved significantly.