摘要： 提出一种用于人脸生理解剖的拟人Agent面部运动模型。将拟人Agent的面部分解为脸、嘴、眼、下巴、舌和头，根据面部辐射肌和口轮匝肌等肌肉的运动特点，通过肌肉变形，推导出控制面部运动的特征参数和辅助参数，由此提出多维的面部运动形式化描述方法。实验结果表明，该方法能精确控制面部的连续运动，准确描述面部表情。

关键词: 面部运动, 拟人Agent, 形式化描述, 人脸表情

Abstract: To construct a uniform model for facial action of human-like Agent, the face is decomposed to six parts which are face, mouth, eyes, chin, tongue and head, according to physical anatomy in the paper. Then the method of multi-dimensional vector description is proposed through analyzing the movements of muscles on the face such as radiation muscles, orbicularis muscle and so on. The multi-dimension vector consists of a set of control parameters of facial action. The priority of the description can re-orientate a facial movement to another facial model. Experimental results show that the method can produce and descript the movement of human face realistically and easily.

Key words: facial action, human-like Agent, formal description, facial expression

中图分类号: 

• TP301