摘要
通过对人脸主要特征位置的粗标定,得到眼睛位置的精标定,进而确定嘴的粗位置,构造了一个嘴的变形模型,通过多变量寻优算法获得比较准确的发音嘴形特征. 这对于人机自然交互,听觉视觉双模态语音识别,计算机视觉的研究都有重要意义. 这种方法简单、快捷,通过试验验证有较好的效果.
This chapter locates the eyes accurately by locating the main facial features roughly and gives a rough estimate of mouth. . For locating mouths accurately , a deformable template is constructed. At the end, the mouth features of speaker are gained accurately by minimize a n-dimensional function . It is important for the study of man machine natural interaction ,audiovisual bimodal speech recognition and computer vision . The algorithm of this chapter is simple and quick. Computer simulation shows the result is very nice.
出处
《五邑大学学报（自然科学版）》
CAS
2002年第1期11-16,共6页
Journal of Wuyi University(Natural Science Edition)
基金
广东省自然科学基金资助项目(000872)
