Abstract: In order to improve the credibility of the human animation, personality model, emotion model and mood model, are designed to represent personal characteristic and emotional states. And a reformative method for computing perception weights is proposed. Among the three models, virtual human's perception at the virtual scene is determined by the personality model and the spatial position, and it also changes the avatar's emotion and mood models, while the changed emotion and mood models alters the behaviors of the virtual human. The updated emotional state motivates the virtual human to react to the scene. Applying this modeling technology and related mapping method to virtual reality system is helpful to synthesizing new motions by calculating emotion model rapidly. In the final animation effects, the authors can find out the influences of spatial position on emotion update, emotion to behavior change and the diversification of different avatar.