Research On Speaker Recognition Technology Based On Wavelet Analysis

Speaker recognition technology is one of the main application of the voice signal processing technology. Now, speaker recognition technology is being studied by many institution at home and abroad, and has become a hotspot in voice recognition area.This paper focuses on speaker recognition technology based on the wavelet analysis. First, the basic principle of speaker recognition is discussed in detail, which relates to speech signal preprocessing, feature parameters extraction, as well as the recognition based on feature parameters. The de-noising method based wavelet modulus maximum threshold and wavelet-domain threshold is analyzed focused, is improved.Besides, two kinds of feature parameters are focused analyzed, that is linear prediction cepstral coefficients-LPCC and Mel cepstral coefficients-MFCC. The multi-wavelets analysis is applied to the extraction and analysis of feature parameters. Experimental result shows that the recognition rate is done due to the analysis based on multi-wavelet. Specific tasks are as follows:(1) Analyzed of the current development of speaker recognition, the technical difficulties and the need to improve on, and researched the associated content of the voice signal processing.(2) Discussed de-noising method based on wavelet modulus maxima and wavelet-domain threshold. Combined both together and improved the de-noising method of wavelet-domain threshold to meet its shortfall with the advantage of wavelet modulus maxima. Experiments show that the effect of speech signal de-noising increased with the improved method of wavelet-domain threshold.(3) Analyzed and integrated the two kinds of feature parameters, while the application of multi-wavelet analysis. Finally, the MATLAB simulation verified the recognition effectiveness with multi-wavelet analysis. In the end, a conclusion of this paper and prospect of the future work are drawn.