Uyghur Voice Phoneme-Viseme Mapping Study

Uighur speech-viseme mapping research is an important part Uighur visual speech synthesis. When given voice, if we can obtain the corresponding visual parameters,which is the Uyghur speech-viseme mapping research to achieve the effect.Uighur speech-viseme mapping methods have a lot of, mainly the method of vector quantization, neural networks, Gaussian mixture model and hidden Markov model,Bayesian dynamic model, etc. Hidden Markov model for computing power is strong,and become the mainstream of the research methods.After analyzing the advantages and disadvantages of each model, the use of hidden Markov model Uighur speech-viseme mapping. The main contents are as follows:First, the comparative advantages and disadvantages of each individual face model, select Candide-3 model, by deforming method to give a specific face model.Second, the reference MPEG-4 standard, to extract characteristic parameters lip phonemes.Third, analyze of characteristic phonemes Uighur, Uyghur speech feature extraction.Fourth, analyze the advantages and disadvantages of audio and video model,introduced algorithm HMM model using HMM model Uighur speech bits- lip mapping.