Algorithm Based On Hidden Markov Shanghai Urban Road Speech Recognition Method

With the rapid progress of modern computer technology and speech technology, communicating through speech and computer becomes a hallmark of the development of computer technology. Accordingly, speech recognition and synthesis become a research field that need to develop urgently.The technology of speech recognition concerns different fields and the development of these fields has made contribution for the development of speech recognition. The characteristics of speech signal have caused the difficulty of speech recognition, these characteristics include changefulness, dynamism, instantance and continuity etc. The process of computer recognition for speech and the identification process for speech by a person are basically consistent. Now, the main stream of the technology of speech recognition is based on statistical pattern recognition.This paper is with the purpose of realizing a city road name speech recognition system based on HMM model. This paper has introduced the thought as well as the improvement in the algorithm of speech recognition in realizing high recognition rate in detail. This paper discussed the feature vector extraction methed, the realization of Baum_Welch algorithm and the principle of HMM method. On this foundation, the author adopted Viterbi algorithm and Segmental K-means algorithm in training HMM model, which reduced operational complexity , raised recognition speed , making it satisfy the requirement of real time execution. At the same time according to the low recognition rate in speech recognition system, the author used the method of fundamental frequency analysis to build male/female recognition model respectively. The author also built two kind of models: one is based on the city road name( whole word) and the other is based on the phoneme of mandarin. Finally this paper gives the comparison and analysis on the recognition property of these two models.