The experiments show that DBN model has the better performance than HMM, and DBN model based on viseme can provide the best mouth shape for talking head.

实验表明，DBN模性能优于HMM，而基于素的DBN模能为说部动画提供最好的口形。