Abstract

We present a method to generate mouth animations coupled with moderate movements of the head. Our method uses a coarse 2D textured mesh representation of images. With hardware texture mapping, we obtain 30 frames/sec animations on modest PC platforms. The synthesized animation sequence is realistic and lively. Radial basis functions are used in the underlying animation engine. At each phoneme transition, the positions of the mesh nodes are determined by RBF coefficients associated with a set of sparse feature points. Blending textures from key frame images on the displaced mesh produces animation sequences of talking faces.

Citations