摘要： 使用改进的基于图像库方法HMM方法进行人脸语音动画合成过程中,为了保证结果视频流的平滑真实,在相应音素的图像帧之间进行插值产生中间帧就成为语音动画中十分重要的部分。该文介绍了从相应语音的图像帧中提取特征区域进而对其中的特征点进行定位,从而将特征区域根据需要进行划分并在前后对应音素的图像帧中进行变形进而得到流畅自然的语音动画效果的工作模块。

关键词: 特征点提取, 图像变形, 语音动画

Abstract: During using the improved method HMM of “based on image-warehouse” to synthesize speech animation, in order to guarantee the frequency of the final result, it must interpolate some frames between frames of the two corresponding phonemes. This paper introduces the method of picking up feature points from correlation frames of phonemes, dividing the feature areas with these points and interpolating some frames between existing frames to get a fluent video-flow.

中图分类号: 

• TP391.41