Research Of Speaker Recognition System Based On VQ

Speaker recognition is the process of automatically recognizing who is speaking on the basis of individual information include in speech signals. It has well application prospects in many fields. In this speaker recognition system contains two main modules: feature extraction and feature matching. Feature extraction is the process that extracts a small amount of data from the voice signal that can later be used to represent each speaker. Feature matching involves two distinguish phases. The first one is referred to the training phase while the second one is referred to as the testing phase.Firstly, in this paper, I analyze the common principle of the voice signal, and focus on the double limit method of measuring the extreme point of the voice signal and the method of measuring the LPCMFCC extreme point. Then I research the three submodel of digital model of voice signal: inspired model, sound gate model, and radiant model. Secondly, based on the theory of feature parameter, I study the parameter representing the speech signal for the speaker — Mel Frequency Cepstrum Coefficients (MFCC) and implement the process of the extraction of Mel Frequency Cepstrum Coefficients.Thirdly, I explain the principle of Vector Quantization (VQ) and the measure of VQ distortion, and illuminate the best arithmetic to build a speaker-specific VQ codebook-LBG algorithm Moreover, in the progress of clustering the Training Vectors, I use the the nearest dividing algorithm to improve the efficiency of search codeword, and to increase the speed of speaker recognition.In this speaker recognition system , I drew the characteristic parameter of the introduction pronunciation by through MATLAB speech processing toolbox , and use MFCC as the parameter of the speech signal, improving discernment performance of system. I adopt LBG algorithm to design the codebook, and design one VQ codebook for each speaker to avoid the quantization error caused by using the one codebook in which all speakers save their feature vectors. This system achieves the high rate of discernment but the low mistake rate, increases the speed of operation, and reduces the amount of calculation...