Research on effect of frequency band energy leakage to wavelet denoising

Using usual threshold selection methods, a problem exists in the wavelet denoising of the signal that contains some high frequency components. When the signal is denoised, the effect of the frequency band energy leakage (FBEL) of the wavelet analysis results in the increase of the noise estimate of the wavelet denoising and makes the wavelet threshold denoising exist an error. The useful information is badly lost in the wavelet denoising process. This makes against the signal feature extraction. The reason of the above problem is opened out. The rules of the noise estimate affected by the FBEL are analyzed, and the necessity of eliminating the effect of the FBEL is shown. Finally, two effective methods are put forward. One method is increasing the sampling frequency and the layer number of the wavelet analysis. The other one is adjusting the correcting factor of the noise estimate. It is shown that using the two methods the effects of the FBEL to the wavelet denoising are well eliminated.