Speech Enhancement Based On Modified Wavelet Threshold Function

With the rapid development of computer and digital signal processing technology,voice signal processing technology constantly imprvoed and matured,and used in real life through the video conference?mobile phone?network video and other forms.While,various noise will affect the effect of voice communciations equipment,serious hinder life of people and normal work.Therefore,it is necessary to solving the noise interference effectively,and enhancing the quality of the voice signal,case more and more domestic and foreign experts and scholars come into this research.And there have been a lot of effective voice enhancment algorithms.the good effect under the condition of high input SNR,while the input of low SNR,the effect is not ideal.So,the rearch of this paper based on various of input SNR condition.We have done the following works:1:Based on the analysis the theory of wavelet denoising,mainly research wavelet threshold function denoising algorithms,the purpose of it were effective for remove noise wavelet coefficients and preserve the useful speech coefficients from noise speech signals,some of existing wavelet threshold functions have various deficiencies in discontinuity and different decomposition level about the fixed threshold aspect,which also produce constant errors during the de-noising process.An improved wavelet threshold function w-ith adjusted parameters is proposed,the particle swarm optimization(PSO)algorithm is implented to find the optimal parameter values of this threshold function in a noise background environment,and the improved wavelet threshold function method combine with Bayes shrink threshold method,furthermore,the processed optimal wavelet coefficients were reconstruct the enhancement speech signal.Compared with the traditional threshold de-noising method,the simulation results show that the modified threshold function can improve the output signal to noise ration(SNR),reduce the useful signal distortion and eliminate background noise effectively.2:The improved threshold function of wavelet denoising algorithm effect is obvious under the condition of high input SNR,but there is residual under low input SNR,therefore,in order to improve the effect of denoising algorithm further,the enhanced voice from the method improved threshold function of wavelet denoising as a priori information,and combine with Kalman filtering speech enhancement method.Many simulation experiments have been done on MATLAB platform,the simulation results show that the speech enhancement based on improved wavelet threshold function and kalman filtering method can achieve better enhancement voice in low input SNR.3:The preparation of the speech signal acquisition and display,which main on the PC windows operation system and based on VC++ 6.0 softwave,through the USB sound card and the audio Wave class provided by Windows to complete audio programming in sound card.Phonetic acquisition interface.the display waveform data function through the drawing API functions provided by Windows.