一种小波自适应阈值全频降噪方法

An Adaptive Wavelet Threshold De-Nosing both in Low and High Frequency Domains

目前的基于小波阈值降噪方法往往假设信号的噪声分布在高频段,因此大部分方法只对高频段进行降噪,而忽略了低频段噪声对信号的影响.在现实的应用中,复杂的噪声并不满足该假设条件,也即复杂噪声不仅分布在信号的高频段,而且低频段的噪声同样不容忽视.针对上述问题,论文提出了一种全新的解决方案：小波自适应阈值全频降噪方法.在该方法中,根据不同类型的噪声随小波分解层数、噪声强度等因素变化规律,提出了一种新的自适应阈值确定方法;然后利用小波去相关性方法来检测信号受到的最主要的噪声干扰;最后结合噪声类型检测方法,检测信号中所隐含的最接近的噪声类型,选取合适的阈值确定方法,对信号的低频和高频同时进行降噪.论文的实验结果表明：（1）当信噪比较低时,采用全频降噪方法对大部分类型的噪声而言均优于传统方法,并且全频降噪方法仅需要信号分解到1-2层即可取得良好效果;（2）当信噪比较高时,全频阈值降噪技术的降噪效果和传统方法一致,但所需小波的分解层数少于传统方法.

It always tends to assume that the noise contained in signal spread over high frequency domain in the traditional wavelet threshold de-noising techniques.However,it doesn＇t hold for different noise categories,and threshold de-noising methods in most literatures rarely mention the noise influence spread over low frequency domain.Thus,a new framework for noise reduction base on full frequency domain using wavelet decomposition and noise-type detection are proposed.In this framework,the noise type is firstly to be detected by analyzing autocorrelation coefficient for different noise,and then noise reduction is performed both in low and high frequency domain.The experimental results show that：（1） when signal-to-noise ratio is low,our method not only always achieves better de-nosing performance,but needs fewer decomposition layers than the traditional methods;（2） when the signal-tonoise ratio is high,our method can obtain the same performance as the traditional methods,but our method needs less decomposition layers.