Study On Speech Enhancement And Denoising Based On Multi Channel

The speech signal is easily disturbed by the noise in the environment.In order to obtain a relatively pure target signal, the speech enhancement method is usually used to denoise the received signal.Generally speaking, the single microphone speech enhancement method has better denoising ability, but its ability to suppress the noise in complex environment is poor.Microphone array has the characteristics of spatial selection and signal gain, which can overcome the disadvantages of single microphone, and has better performance in signal gain and anti-jamming.In order to get better denoising effect, this paper uses multi-channel speech denoising method to denoise speech signal with noise:(1)In this paper, the fixed blocking matrix in the generalized sidelobe canceller can not block the target signal well, and has a limited ability to eliminate the incoherent and weak coherent noise. An improved GSC speech enhancement method is proposed, which uses the adaptive blocking matrix to replace the fixed blocking matrix in GSC to track the target signal continuously and reduce the loss of the target signal.(2)Generally speaking, it is difficult to use the soft threshold and hard threshold to denoise the speech signal in wavelet transform.Put forward an improved method of wavelet transform, the method using threshold function and improved adaptive threshold for decomposition of wavelet coefficients and wavelet coefficients of the smaller amplitude of the reconstructed extracted noise for RLS adaptive filtering for follow-up, so as to reduce the impact on the target signal denoising process.(3)In view of the problem that the RLS adaptive filter can be affected by the change of the system when the system changes, the RLS adaptive filtering system is proposed.RLS adaptive filtering system of this system in the traditional variable forgetting factor is added in a correction function, after this function at the iteration of forgetting factor is modified, not only successfully solve the tracking ability of RLS adaptive filtering system, the problem of low stability, but also solve the high requirements on the adaptive algorithm the calculation accuracy of the equipment problems.(4)The improved GSC and wavelet transform and improved RLS adaptive filter three is combined to propose a new denoising algorithm, the algorithm can not only in the spatial range of the target signal and spatial filtering in time domain and frequency domain, the wavelet domain of the residual noise and noise reduction the further processing of the target signal, thus obtaining the ideal.