基于交互式进化计算的虚拟表情建模

VIRTUAL FACIAL EXPRESSION MODELING BASED ON INTERACTIVE EVOLUTIONARY COMPUTATION

探讨如何训练虚拟人物表情这一新课题,提出了将TSK模糊神经网络应用于虚拟人物表情的建模研究,并用交互式进化计算对TSK模糊神经网络进行训练。实验结果表明,当该算法应用于虚拟人物表情建模时,比传统的进化计算具有更快的收敛速度,而且该算法融合了进化计算的全局优化和人的主观判断,使得表情输出更加符合用户的需要.从而使得不同的动画导演利用这个系统,能够根据自己的知识产生出符合自己要求的虚拟表情输出。

Facial expression modeling of virtual actor is an important and practical topic. TSK fuzzy neural network is applied to accommodate director＇s expertise so that appropriate facial expression of virtual actor can be produced. In order to speed up the convergence of the learning algorithm ,Interactive EC （IEC） algorithm is incorporated into TSK fuzzy neural network. IEC fuses the capability of EC optimization and the human evaluation. Experimental results indicate the success of the approach.