The Study Of Improved GSC Algorithm Based On Microphone Array Speech Enhancement

The purpose of speech enhancement technique is to minimize the noise and extract the pure speech signal from the noise signal.The noise reduction algorithm of single channel microphone voice signal has a good noise reduction effect in the high signal-to-noise ratio(SNR)condition.But output signals degrade significantly in the low SNR condition,which leads to the low identification of the signal.Later utilizing the microphone array is proposed to make speech enhancement.The desired speech signal is received as the beam is formed to aim at the speech signal,i.e.,we can use receive beamforming technique to make the gain of main lobe towards the desired signal.Therefore,noise signal can be effectively eliminated.In this paper,we focus on the generalized side-lobe cancellation algorithm(GSC).Firstly,acoustic environment algorithm is investigated according to the received signal of microphone array in different acoustic environments.In the simple acoustic environment,we elaborate the conventional microphone array speech enhancement algorithm based on the structure of GSC,then using the independent component analysis method to improve the conventional algorithm.The improved algorithm has a significant effect on speech enhancement.In the complex acoustic condition,utilizing the improved algorithm has low effect of the speech enhancement.Therefore,TF-GSC algorithm is proposed.This algorithm has a great effect of eliminating interference noise for the directionality noise,which leads to residual noise of output signal.Meanwhile,the Chapter 5 illustrates wiener filtering algorithm based on wavelet to further improve the TF-GSC algorithm to filter out the residual noise.We collect the speech data to make the simulation analysis,which confirm that the improved TF-GSC algorithm can effectively eliminate the noise spectrum of the speech signal,and the algorithm is a less signal distortion.Then,we find that the direction estimation error of the desired signal will produce great impact on the performance of the algorithm.In order to solve this problem,Chapter 6 illustrates a new GSC algorithm based on double affine structure(DAP-GSC),the simulation results show that DAP-GSC algorithm allows for 10° direction estimation error of the desired signal.DAP-GSC outperforms the conventional algorithm in output SNR in the same simulation conditions.