三维可视语音合成系统中唇部特征点的采集与处理

Acquisition and Processing of Feature Points of Lips in Three-dimensional Visual Speech Synthesis

对可视语音合成中的三维唇部特征进行分析,并利用三维动态捕捉系统完成了特征点确定及特征点运动数据的采集工作及三维唇部特征点的选择和采集。首先对唇部特征进行分析,从而提取唇部主要特征点,借鉴MPEG-4对人脸特征点的规定,按照对称性特点的要求来对上述情况进行多次处理,经过反复标注实验,最终确定了本研究的唇部特征点。主要利用三维动态捕捉系统,对唇部特征点数据进行采集,对采集后得到的数据进行校正处理。本文主要讨论三维可视语音系统中三维数据的采集以及处理方法。

This paper analyzes the three-dimensional characteristics of lips in visual speech synthesis. It uses three-dimensional dynam- ic capture system to complete the works of determining feature points, the collection of motion data of their feature points, as well as the selection and acquisition. This chapter analyzes the characteristics of lips, and then extracts the main feature points of lips. By drawing references from the provisions of MPEG-4 on facial feature points, according to the requirement of symmetry, after repeated tagging ex- periments, the lip feature points are determined in this study. This study mainly uses the three-dimensional motion capture system to collect the lip feature points data, and then makes data correction processing. This article focuses on 3D data acquisition and process- ing method in 3D visual speech system.