Research On Microphone Array Speech Enhancement Method

With the popularity of smart devices,the voice communication has become an indispensable part of people's lives.The processing of speech signal requires high quality of the speech signal.However,the picked-up speech signal has a noise signal,so that the clarity and intelligibility of the speech signal are seriously degraded in the case of complicated use environment of the speech system.Therefore,some speech enhancement methods must be used to denoise the speech signal.Microphone array speech enhancement technology can use the spatial selectivity of the spatial filter to track the moving target sound source in real time to obtain high quality speech signals.In this paper,based on the combination of linear adaptive filtering and nonlinear adaptive filtering in the microphone array,the noisy signal is processed.The main research work includes:(1)Aiming at the problem that the generalized sidelobe canceller has poor noise canceling ability for nonlinear noise,a cooperative adaptive filtering speech enhancement method for microphone array is proposed.The method uses a Legendre filter and a linear filter to form a cooperative filter to replace the blocking matrix in the GSC,and synchronously block the target signal.The experimental results show that the cooperative adaptive filtering speech enhancement method of the microphone array can suppress Gaussian noise and also has a good removal effect on impulse noise.(2)The problem of speech noise at each sampling point cannot be accurately estimated for the minimum mean square error algorithm.The combination of LMS linear adaptive algorithm and trigonometric expansion function is proposed.This method has strong removal ability for Gaussian noise,but the suppression effect on impulse noise is not obvious.When the background noise contains impulse noise,the minimum p-order mean square algorithm is replaced by the LMS algorithm.Experiments show that the function link(LMP)adaptive filtering can make up for the lack of function of link(LMS)adaptive filtering on impulse noise suppression.(3)For the problem of small threshold function and fixed threshold function used in wavelet transform,the effect of denoising has a large gap between different noise sources,and an improved wavelet transform method is proposed.The method uses the improved threshold function and the wavelet entropy adaptive threshold to process the wavelet coefficients,remove the noise wavelet coefficients as much as possible,and improve the quality of the speech signal.Combining the function link adaptive filtering method with the improved wavelet transform,a microphone array speech enhancement system based on wavelet transform is proposed.The system can realize the spatial filtering of the target signal in the spatial range and the residual noise in the wavelet domain.Further noise reduction,resulting in an enhanced speech signal.