基于HTK的语音驱动唇形动画的实现  

Application of Speech Driven Lip Animation Based on HTK

作　　者：高春梅[1] 郑伯川[2] 

机构地区：[1]四川大学计算机学院机器智能实验室,成都610064 [2]西华师范大学数学与信息学院,南充637002

出　　处：《现代计算机（中旬刊）》2011年第11期3-7,共5页Modern Computer

基　　金：国家自然科学基金资助项目(60971109)

摘　　要：语音驱动唇形动画的同步是人脸动画的难点之一。首先以音节为识别单位,通过严格的声韵母建模方法,利用HTK工具包,识别得到语音文件中的音节序列与时间信息;然后利用基本唇形库和音节到唇形映射表,获得与音节序列对应的唇形序列;利用唇形序列的时间信息插值播放唇形序列,实现语音驱动的唇形动画。实验表明,该方法不仅大大减少了模型数目,而且能准确识别音节序列以及时间信息,可有效地实现语音与唇动的同步。Lip synchronization is one of the key issues in speech driven face animation system. First se- lectes as recognition unit. Through a method based on rigorous initial and final modeling, uses HTK syllable to get syllable list and time information. Then uses the lip library and syllable-lip mapping table to obtain lip sequence corresponding with syllable sequence. Finally, speech driven face animation is achieved by interpolating and playing lip sequence with time informa- tion. Experimental results show that this method not only sharply reduces the number of models, but also recognizes and plots out the syllable and time information accurately in speech files. It effectively realizes the synchronization of speech and lip animation.

关 键 词：语音识别 语音驱动 唇形动画 声学模型 HTK 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]