Study Of Technique For Environmental Noise Reduction In Speech Signal Based On DSP

The communication difficulties result from the tremendous noise on the battlefield is a very serious problem.There are variety of speech de-noising methods in China,but they are almost all under de-noising processing in laboratory conditions for the smooth Gaussian white noise.As in strong noise envirnment,it is always unsatisfactory.There are also a small number of de-noising methods work well with huge amount of computing or realization of the difficulties.In this paper,several of current major speech enhancement methods such as Wiener filtering,Kalman filtering,spectral subtraction are studied and then simulated in matlab.And they have been done a comparison in the realization and actual de-noising effect. Then base on a priori SNR improve spectral subtraction to do a variety of improvements to better distinguish between voice and noise.Denoising results have been improved once again.At last,I has successfully achieved real-time voice signal acquisition,noise reduction processing and playback through TMS320VC5402.