Abstract: In practical engineering applications, the signals collected by physical methods have noise information and this will submerge many useful signals used for analyzing the system characteristics. The traditional filtering techniques, such as band pass, low pass, high pass, seems powerless, so extracting useful feature information requires denoising of the original signal. Common wavelet threshold denoising methods have many shortcomings. In this paper, a threshold function is re-constructed on the basis of wavelet threshold denoising methods. To realize the simulation of denoising the Gaussian white noise, in the Mtalab (2014a) environment, the conventional hard, soft threshold function and the new threshold function are respectively used. Results show that signals become better and clearer by using the new threshold function.