Research Of Voiceprint Recognition System

Just as fingerprint, face, voiceprint is a kind of biological characteristics ofthe typical human. Because of its convenience and economic, voiceprintrecognition as one of biometric authentication technologies step into the publiceyes gradually. Clearly voiceprint recognition is different from speechrecognition, it pays attention to the speaker’s voice features in the voice signal,While, there is no need to know the mean of words. Because of everyone has theunique voiceprint feature, it is difficult to be imitated and forged, Therefore,compared with other biometric authentication technology, voiceprint recognitiontechnology is more safe and reliable in the field of identity authentication.This article is to do research on the key technology of text independentvoiceprint recognition system, and tries to make some innovations on the basis ofprevious studies, in order to improve the recognition rate of the system. Firstly,this article studies the research background and prospect of the voiceprintrecognition system together with the basic structure and principle and analyzesthe part of endpoint detection, Secondly, does a detailed study of featureextraction. In this part, this article analyzes the extraction process of LPCC andMFCC, and does some improvement of MFCC and raises the algorithm ofcombination feature based on Wavelet Transform and improved MFCC. Lastly,compared with different algorithms of feature extraction by using the Gaussianmixture model by means of experiments, These experiments show that thesemethods improve the recognition rate of the system.