The Research Of Speaker Recognition

Speaker recognition is the main applications of speech processing techniques and one of hotspot at present. Vector quantization and Wavelete analysis theory are introducedin this paper, some workings are as follows:First, based on VC++6.0, a real time speech data sampling program is developed and provide real time data for the research of "Speech recognition"and "Speaker recognition".Second, VQ and improvemental VQ arithmetic is presented in this paper. Model built for the speaker considers frame character of speaker about subtly probality distributingExperiment show that this method improved recognition rata.The last,the principle of wavelet based speech improvement is expatiated and arithmetic is presented.Experiment show that the arithmetic not only has excellent effects on speech pre-processing but also has potential to improve robustness of a speaker recogniyion system in noisy environments.