Speech Enhancement Algorithm Of Non-stationary Noise Environment Research

Speech enhancement is an important branch of speech signal processing. It is also an important pre-pretreatment technology of speech recognition, speech compression coding and speech synthesis. The main purpose of speech enhancement is extracting the pure primitive voice as far as possible from the speech signal with noise and improving speech intelligibility. This paper do research on key steps and denoising algorithm in speech enhancement based on analysis of existing speech enhancement algorithm which used short-time spectral estimation. The main research work as follows:1This paper proposed an improved MCRA algorithm based on MCRA noise estimation algorithm to overcome the disadvantages of the traditional VAD noise estimation algorithm this can only adapt to the stationary noise environment. The improved MCRA algorithm is a kind of adaptive better noise estimation method in non-stationary noise environment which does not depend on the fixed time window to search the local minimum value of noisy speech power spectrum and overcome the noise estimation delay problem of the MCRA algorithm.2This paper analysis wavelet de-noising method and do research on design threshold function in wavelet threshold de-noising algorithm. An improved threshold function design on the basis of non-negative dead zone threshold function which not only has good continuity but also overcome the lack of the fixed deviation existence in the soft threshold function and considers the characteristics of the attenuation of the noise wavelet modulus values conform exponentially. Simulation results show that the improved threshold function can effectively eliminate the noise and gain better enhancement.3In order to eliminate the background noise and improve voice quality in the maximum extent, the crude estimated of enhance speech can get from improvement spectral subtraction which spectral subtraction parameters ware adaptive adjustment by combined human auditory masking effect and spectral subtraction. The optimal smoothing factor based on MMSE criterion designed to eliminate the residual musical noise further.The simulation results show that this method can suppress the musical noise effectively and improve signal to noise ratio at the same time maintain good speech intelligibility.