基于最大似然估计的小波阈值消噪技术及信号特征提取

Wavelet De-noising Based on Maximum Likelihood Estimation and Its Application for Feature Extraction

小波阈值消噪技术是近十年来发展起来的一个新方法,它因具有强大的去噪能力而被迅速应用在许多领域。针对工程中常见的具有稀疏概率密度形式的信号,推出了基于最大似然估计准则的小波消噪方法,并以脉冲信号为例,通过与现有的小波阈值消噪作比较,证实了该方法的优越性。最后,将该方法用在识别齿轮裂纹特征,收到了很好的效果。

Signals with sparse probability density distribution are often encountered in engineering. However, current wavelet de-noising methods do not work well for them. A new wavelet de-noising method is proposed to remove noise for signals with sparse probability density distribution. It utilizes the prior knowledge about the probability density distribution of the signal,and uses maximum likelihood estimation to derive the thresholding rule. This method shows excellent performance when applied to simulated impulses through comparison with current wavelet de-noising methods. The efficiency of this de-noising method is also verified by its application to extract early fault signature for a gearbox.