Research on Non-stationary Acoustic Signal Filtering Method Based on Wavelet Threshold Denoising

In order to deal with the noise problem in the process of acquiring sound information, wavelet analysis method is used in this paper: (1) The traditional wavelet analysis method has the problems of "overkill" and large deviation of the reconstructed signal. By analyzing the wavelet decomposition process and fitting the change of the noise with the decomposition layer, a dynamic threshold calculation method is proposed, which optimized the signal excessive filtering problem; (2) A new threshold function is designed by fusing the characteristics of the soft threshold function and the hard threshold function. The characteristic curve presents a smooth and gradually increasing characteristic of concave function, which solves the problem of large deviations in reconstructed signals. The experimental comparison of signal-to-noise ratio and root mean square error shows that the designed method in this paper can improve the accuracy of signal recognition and has higher application value in processing one-dimensional non-stationary signals.