Construction and animation of realistic human facial models is an important research field in computer graphics. How to simulate the motions of human faces on 3D facial models in real-time to generate realistic facial expressions is still a challenge. In this paper, a technique to simulate the human facial animation realistically in real-time is presented. First of all, the 3D facial model is divided into independent parts named functional regions according to facial motion characteristic. Then the motion of each functional region is simulated by a hybrid method based on weighted Dirichlete free-form deformation (DFFD) and rigid body motion simulation. Intercrossed control points are designed to simulate motion affection between different functional regions. The animation of facial models is driven through control points. In order to drive the animation of facial models more efficiently, two high-level methods are proposed, one is based on FAP (facial animation parameters) stream of MPEG-4 standard, and the other is based on muscle model. With these methods, highly realistic facial animation can be achieved in an efficient way.