Research On Text-driven Visual Speech Synthesis Technology

With the rapid development of computer graphics and multimedia technology in the past few years, visual speech computer animation of highly realistic has become a research hotspot in computer science field, which is widely used in deaf education, electronic affairs, movie stunt, human-computer interface, medical surgery and other aspects.This thesis focus on the study of text-driven visual speech synthesis technology, with the purpose of analyzing the characteristic of the inputting text information, extracting out of voice, facial expressions and timing control information.. A real sense of naturalness speech synchronized facial animation will be simulated by making use of the improved 3D human face model .Firstly, deep analysis of the existing methods of 3D face modeling has improved the specialized facial model to reduce grade points and the number of editable mesh, thereby decreasing the computational complexity and the alleviating the system pressure .Secondly, by emphatically analyzing physiological features of human faces, summarize the speech process facial muscles, and propose muscle abstraction methods. This method is to simulate the key parts of the face through grid model, in order to conquer the defects of 3D human face model deformation and mesh vertices beyond our control weakness in the existing methods. Thirdly, according to the characteristics of the inputting text information research , put forward a method of embedding expression tags to provide expressions and time information for subsequent synthesis of facial animation. In addition, according to mandarin pronunciation rules for estimation of each word in the flow of language long pronunciation, as synchronous control condition of visual speech animation. Moreover, analyze the degree of mutual influence of vowel and consonant changes in mouth shape in a continuous speech flow, then grade them. And to improve the Chinese pronunciation model, supported by inter-frame integration of transition treatment, synthesize speech synchronized facial animation. Realize the text-driven visual speech synthesis system.In the last place, through the related experiments on the face/embouchure frame fusion transition handle, realize of the visual speech text drivers face animation. Experiments have proved that this model, with high practical value, can simulate the process of face expression changing truly and naturally.