Research On3D Text-Visual Synthesis Technology

This paper studies a text-driven3D lip visual speech synthesis technology. Firstly this paper analyzes the three-dimensional characteristics of the lip, using three-dimensional motion capture system to determine the feature points, and collect the motion data of the feature point.With the method of image stitching sequence we can synthesis3D visual-speech system. This approaches combines the key frame deformation method, and also the confirmation and extracting of the key frames. The key is extracted out of the frame as a frame-based image stitching.it is used for the synthesis of transition-frame image stitching sequence. In the analysis of the inputting text using the C++programming, as the characters-Pinyin library is already exist.we use the classification methods of Chinese phonemes and visual phoneme to convert the enter text to the corresponding pinyin.and then split it into the smallest phonetic phoneme. For the approach of text-driven. First determine the visual phoneme classification based on s.Then determine the mouth shape in three-dimensional mapping betweenmouth phoneme variation. Using this mapping to establish phonetict-shape mapping.At last create a text-shape mapping.On the basis of these studies, by entering the text can obtained the corresponding three-dimensional mouths.According to the image mosaic method and interpolation method in3D image, the output image sequence can reflect the state of the lip movement much truly. Software simulation shows that the method of this research achieves text-visual speech synthesis on a certain degree. Compared the out put with the the lip motion effects and real-time three-dimensional motion capture system.THE output images of Text-Visual system have no jump points.and the motion process is more clearer.