Signal de-noising is a very important engineering problems, and hasa range of applications in communication, machinery,medicine etc. Signal de-noising has along development history. Along with the mature of wavelet theory, wavelet analysis takes the place of fourier analysis in signal noise reduction processing gradually. Wavelet analysis is a time-frequency analysis method,and has good frequency resolution in the low frequency part of the signal, and has good time resolution in the high frequency part of the signal. Wavelet analysis is known as mathematical microscope.First the article analyzes the general traditional threshold, the universal threshold has a better effect in the single denoising, but for vibration signal, with the increase in the level of decomposition, the choosing the same threshold of each layer has some shortcomings. In order to solve this problem,. A new adaptive threshold which based on the traditional threshold is put forward, the new threshold can choose adaptive threshold according to the different decomposition levels of layer.Secondly, for the traditional threshold function, Hard threshold function has a good Performance in retaining the Partial features of signal edge and in denoising mutant signals.Soft threshold funetion is more smooth, however could lead to distortion such as vague signal edge.To balance the strength of hard threshold andsoft, On the basis of traditional threshold function,the paper propose a new double threshold function which applied to the vibration signal denoising he new threshold function has a smooth transition between the noise and signal, which make it more consistent with the continuity of the vibration signal.Finally, the paper proposed an adaptive lifting wavelet based on the lifting wavelet which appling on the denoising.the prediction funtion of the a adaptive wavelet has a algorithm of nonlinear appling on the sequence. In the practical application, the vibration signals are usually non-stationary and nonlinear, linear prediction mode will inevitably bring unnecessary errors, the algorithm of nonlinear could be as close as possible the more primitive signal. |