The Research Of Associated De-noising Method For The Voice Based On DSP

Because of the non-stationary and the complexity of speech signal, the desired noise reduction targets difficult to be achieved for the traditional method, especially for the broadband noise suppression. An associated de-noising method for the voice was presented on the basis of FIR and wavelet threshold de-noising in this paper; the noise outside band of voice signal was filtered by the FIR filter of optimization, On the basis of this, the original signal was efficiently restored by wavelet threshold de-noising, and realized in DSP.Firstly, the efficient algorithm--Remez algorithm based on alternate theorem was introduced for the insufficient of FIR common design method, it was used to optimize the FIR filters to improve the accuracy of design. Based on the performance comparison of differnt design methods, the simulation results showed that the Remez algorithm based on the optimization design method can offset window function and frequency sampling method, proved Remez optimal filter algorithm can achieve the ideal effect of filter by actual instance.Secondly, the wavelet threshold de-noising method was introduced on the basis of Mallat algorithm for the insufficient of common noise reduction method. Wavelet threshold de-noising theory and the theory of analysis of noise steps were expounded, the denoising performance of soft threshold, hard threshold and soft and hard thresholding eclectic method were analysed in theoretics. MATLAB simulation experiment proved that the original signal was restored by wavelet threshold de-noising in different SNR, but the noise reduction result was better in higher SNR.In the environmental CCS of DSP, the combined de-noising method for the voice based on FIR and wavelet threshold de-noising was simulated. The program design ideas of all noise modules and configuration of important chip were discussed in detail, the main code were given. Finally, the combined noise reduction for noise speech were tested on the basis of C5402 development board, the experimental results shew that noise reduction effect was remarkable with the associated de-noising method for the voice on the basis of FIR and wavelet threshold de-noising ,it has good application prospects.