Research On Multichannel Speech Enhancement Algorithm

In real life,voice signals are inevitably affected by environmental noise,resulting in a sharp deterioration in the performance of many voice processing systems.The microphone array has spatial selection characteristics and high signal gain characteristics,which can get more voice information,so that you can better distinguish between voice and noise signals.However,a single microphone array speech enhancement algorithm is also difficult to meet the requirements,in recent years,based on the microphone array of speech enhancement methods and other algorithms combined method has gradually become a speech enhancement technology research hotspot.This paper also studies the speech enhancement algorithm based on the combination of microphone array.The main work is as follows:Aiming at the problem of noise spectrum estimation in traditional Kalman filter,an improved method based on energy histogram noise power spectrum estimation is proposed.The basic method is based on the histogram of the power spectrum of the past to obtain the noise spectrum estimation,that is,for each new frame,first in a few hundred milliseconds time window to build a power spectral value of the histogram.The power value corresponding to the maximum value in the histogram is used instead of the noise power spectral value.The simulation results are verified by two simulation experiments.Experiment 1 is the minimum statisticaltracking of the noise spectrum estimation algorithm and the energy quantity based on the noise spectrum estimation of the improved algorithm for noise spectrum estimation simulation.The simulation results show that the improved noise spectrum estimation method is closer to the real noise.In the second experiment,the improved noise spectrum estimation algorithm is introduced into Kalman filter,and the improved Kalman filter algorithm is simulated and verified.The simulation shows that the speech component of the signal is not canceled.However,the improved Kalman filter still retains more background noise,because the single channel speech enhancement algorithm does not meet the requirements,need to use the multi-microphone array to get more voice information to complete the denoising.Aiming at the problems of improved Kalman filter and the shortcomings of the traditional microphone array speech enhancement algorithm,this paper proposes a method to combine improved Kalman filter with generalized sidelobe canceller.The use of improved Kalman filter as post-filter,to overcome the GSC on the non-coherent noise suppression is not obvious and the residual more noise shortcomings.The simulation results show that compared with the traditional GSC adaptive algorithm and post-filter algorithm,the improved algorithm has smaller background noise and higher output signal to noise ratio.At the same time,subjective and objective evaluation mechanism is used to compare several algorithms,which proves that the improved algorithm based on GSC structure has better denoising performance...