Abstract: The obtainment of speech signal feature is the premises and foundation of audio-visual conversion in speech driven face animation. The obtainment technology of speech signal feature is studied. As for parameter precision, wavelet transform is introduced to reconstruct original speech signal. This research lays the foundations of audio-visual conversion model in the speech driven face animation system.

Key words: Audio-visual conversion, Speech signal feature, Wavelet transform

摘要： 语音特征参数的提取是语音驱动人脸动画中语音可视化的前提和基础，该文立足于语音驱动的人脸动画技术，较为系统地研究了语音参数的提取。在参数精度方面，引入了用小波变换重构原始信号的思想，对重构后的信号进行参数提取，从而为语音驱动人脸动画系统建立良好的可视化映射模型奠定了基础。

关键词: 语音可视化, 语音特征参数, 小波变换