The Research Based On The Text Irrelevant Speaker Distinguishes

With the fast development of information technology ,more and more place needs to check people's identity reliably. The traditional identity authentication in the form of password already exposes a lot of drawbacks. For insuring information security ,the technology of using people's peculiar biological feature as the authentication means develop gradually. Speaker recognition technology belongs to biometrics recognition technologies ,which is the technology of automatic identity speaker's authentication ,according to obtain the feature parameter which reflecting mental and physiology in voice form. Voiceprint recognition become more important and popular way of ensuring security ,because of its convenience, accurate ,economical and well expansibility.This text researches text-independent speaker verification based on VQ. Firstly, the fundamentals of the speaker recognition are discussed in detail. Then the followings are preprocessing and feature extraction procedures. After that are the windowing, noise filtering, end-points detecting of speech signals. The double threshold end-points detecting is discussed in detail, and has given the traditional double threshold vertex examination algorithm procedure. . In view of traditional double threshold vertex examination algorithm compatibility not very well , the text proposed one improvement vertex examination method: energy frequency value vertex examination algorithm. Has analyzed the current most commonly used three kind of phonetic feature parameter emphatically: linear prediction coefficient, linear prediction cepstrum coefficient, mel frequency cepstrum coefficient, and has withdrawn these three kind of coefficients.In addition, the text discussed the speaker to distinguish the commonly used method emphatically: based on vector quantization speaker recognition methods . For VQ algorithm, the definition of VQ concept, distortion measure, and the method for fittest codebook design are firstly fully unwrapped, then delved into fuzzy VQ (FVQ). The experimental result indicated that recognition rate based on the FVQ is good than recognition rate based on the VQOnce more, has compiled take△MFCC+MFCC as the parameter,Based on vector quantization training and test order. To the VQ model, the codebook number's selection to distinguishes rate has the very tremendous influence, The experiment proved selects 128 most appropriate.In the final, conclusion has been derived about the research, and some advice for further researches has also been provided.