基于运动分解与权重函数的嘴部中文语音动画

Chinese speech animation based on decomposition of mouth motion and weighting function

传统的嘴部肌肉建模方法中对肌肉运动划分比较细致与独立,设置的控制参数较多,缺乏对肌肉运动的分解以及运动控制相关性的分析。提出一种基于运动分解的嘴部子运动单元模型。该方法首先建立基于标准Cadide-3模型的嘴部改进网格模型;然后根据解剖结构及肌肉运动分析,将嘴部运动分解为下颌骨旋转、唇部的收缩与舒张、凸出与上翘三个基本运动单元;最后根据输入的文本信息与音素间视觉权重函数,并结合嘴部的子运动合成得到语音同步的唇部动画。实验结果表明,该方法能够快速合成与中文语音相匹配的唇部动画。

Traditional mouth muscle modeling methods divide the muscle movement relatively detailed and independent, set more control parameters,lack of decomposition for muscle movement and analysis of motion control correlation. This paper presented a mouth sub-movement unit model based on motion decomposition. Firstly, the method established an improved mouth model based on the standard Cadide-3 model. Then, according to the anatomical structure and motion analysis, mouth movement would be decomposed into three basic motion units which were mandibular rotation, labial contraction and relaxation, bulge and upwarp. Finally, according to the input text information and visual weighting function between phonemes, and combining with mouth subovement, this unit model can compound and get mouth animation synchronized with speech sounds. The experimental results show that this method compounds lip animation matched with Chinese speech sounds rapidly.