基于Cross-Validation的小波自适应去噪方法  

Adaptive Wavelet Denoising Based on Cross-Validation

作　　者：黄文清[1] 戴瑜兴[1] 李加升[1] 

机构地区：[1]湖南大学电气与信息工程学院,湖南长沙410082

出　　处：《湖南大学学报（自然科学版）》2008年第11期40-43,共4页Journal of Hunan University:Natural Sciences

基　　金：湖南省自然科学基金资助项目(05JJ40001);长沙市科研基金资助项目(K051150-72)

摘　　要：小波去噪算法中,阈值的选择非常关键.提出一种自适应阈值选择算法.该算法先通过Cross-Validation方法将噪声干扰信号分成两个子信号,一个用于阈值处理,一个用作参考信号;再采用最深梯度法来寻求一个最优去噪阈值.仿真和实验结果表明:在均方误差意义上,所提算法去噪效果优于Donoho等提出的VisuShrink和SureShrink两种去噪算法,且不需要带噪信号的任何'先验信息',适应于实际信号去噪处理.A new wavelet-based adaptive denoising algorithm was presented. By using a modified twofold cross-validation, a noise-corrupted signal was divided into two parts： one for estimating, and the other as a reference signal, and they made it possible to search for optimal thresholds by using gradient-based adaptive algo- rithms. The numerical results indicate that the proposed method outperforms the standard wavelet thresholding denoising methods, like Donoho＇ s VisuShrink and SureShrink, in MSE sense. The proposed algorithm does not need any prior information of the noise-distorted signal, and its convergence speed is faster. So it is fit for realtime signal processing.

关 键 词：小波变换 Cross-Validation 自适应滤波 阈值 

分 类 号：TP202[自动化与计算机技术—检测技术与自动化装置]