Text-Dependent Speaker Recognition Algorithm Study Based On Improved VQ

Speaker recognition is a kind of biometrics technology.It extracts speech features from collected speech utterance,trains speaker's model,then determines speaker's identity.Speaker recognition has broad application foreground in many fields such as electric business and information security.By analyzing speech characteristic parameters and the basis methods of speaker recognition,we chose MFCC and LPCC's difference to be the speech characteristic parameters.Using DTW and VQ to recognize text-dependent speech ,we have developed a speaker identification system in this paper.Firstly,this disquisition introduced the foundation acoustics knowledge of speaker recognition:the mathematic model of speech signal and the characteristic parameters of speech both in time-region and frequency-region.By using full pole model,we obtained speech signal LPC,then deduced it's LPCC,and we used the LPCC difference to describe speaker's track dynamic movement.Also,since MFCC represents hearing frequency nonlinear characteristic,we utilized MFCC to be another speaker recognition characteristic parameter to distinguish the input passwords.Next we discussed two kinds of methods of speaker recognition: DTW and VQ.But there's still some defect in the code word bringing of VQ.So,in the process of speech modeling,we added a variance to the code word of VQ and got Continuous Vector Quantization.Then,the code word is represented by a couple of vector,it can show the dispersion extent of the characteristic parameters distribution more clearly.An own base of speech was established according to experiment's requirement practically,which includes text-dependent speech recorded from 30 speakers.Then we utilized MATLAB Voice Box to extract speech's characteristic parameter,used CVQ to matching reference model with test model and compared the method with DTW and VQ in text-dependent speaker recognition experiment.A higher recognition rate was acquired.