Research And Speaker Recognition System

For human being language is the most nature, convenient and efficient way to communicate with each other. Speech is one of the key media to make language spread. With the development of science and technology the interface between human and the machine has experienced three stages: hand-operating interface, text interface and graphic interface. The new generation interface of intelligence is under research. Soon it will become one part of our daily life. One of the essential feature of intelligent human-machine interface is speech communication. So speech recognition has become an active research area. Many countries invest lots of manpower, material resources and finance to solve this hard problem.The focus of this paper is speaker recognition. It is a branch of speech recognition . It can be divided into speaker identification and speaker verification. Firstly the basic theory of speaker recognition is discussed. Considering the performance of the current speaker recognition system degrading sharply with low SNR, wavelet analysis is introduced into speaker recognition research. Traditional speech segment methods and parameters representing speaker identity are improved. In order to make the recognition rapid a recognition model based on searching tree is proposed Lots of experiments on it were done and the results proved what we proposed is possible. Finally a text independent speaker identification system is implemented. It is proved that the theory we introduced is effective to speaker recognition application .