Research On Speaker Recognition In Noisy Environment

Along with the development of social information, more and more situations need the identification of the person and traditional identification of person by using cipher has many disadvantages. In order to ensure information safety, the technology of using people's peculiar biological feature as the authentication has been developed gradually.Speaker recognition technology belongs to biometrics recognition technology,which is the technology of automatic speaker authentication according to obtaining the feature parameter which reflect behaviour and physiology characteristic in voice wave. With the development of computer technology, people have attached importance to speaker recognition technology. The speaker recognition technology has two key aspects: One is how to extract data from many original speech signals which reflect personal characteristic.The other is how to design the valid classifier.Speaker recognition's research has a long history.A speaker recognition system with high recognition performance in relatively clean environments will become deficient in the noisy environments.In order to advance recognition rate in noise environments, extracting combination feature based on wavelet transform GDS is proposed considering the advantage of wavelet transform.As to the classifier,Vector quantization technology is adopted.Expeiments in the different noise and different SNR show that the validity of feature parameter is proved.Firstly, background and application of speaker recognition technology are introduced and then speaker recognition technology are generaliy analized. Secondly, common parameters of speaker recognition are discussed ,and through experiments their robust performance are compared.After that under noisy environment a new combined feature parameter is given and its extracting procedure is introduced in detail.Then classifier method is designed.At last according to this method ,Speaker recognition system is realized.