基于卷积型小波包变换的信号消噪算法  

Denoising Algorithm Based on Convolution Type of Wavelet Packet Transformation

作　　者：赵学智[1] 陈统坚[1] 叶邦彦[1] 彭永红[1] 

机构地区：[1]华南理工大学机械工程学院,广州510640

出　　处：《数据采集与处理》2003年第3期292-295,共4页Journal of Data Acquisition and Processing

基　　金：国家自然科学基金 (5 990 5 0 0 8) ;华南理工大学自然科学基金 (E5 3 0 5 2 92 )资助项目

摘　　要：提出了卷积型小波包变换 ,与传统小波包变换相比 ,在这种小波包变换中不管信号被分解多少层 ,每层分解得到的各频道序列长度始终与原始信号一致 ,利用这一性质本文进一步提出并实现了对小波包分解结果利用模极大值法进行消噪的算法。这一算法的思想来自于基于小波变换的模极大值消噪算法 ,但是由于小波包分解是对小波分解的结果作进一步细分 ,在小波分解中难于分离的高频噪声将被小波包充分分离与集中到后面的频道 ,因此基于小波包变换的模极大值消噪算法将会取得比小波消噪更好的效果。文中给出了信号的小波包消噪实例 ,并与小波消噪的效果进行了对比 ,结果表明小波包有更优良的消噪效果。The convolution type of wavelet packet transformation is put forward. Compared with traditional wavelet packet transformation, in this wavelet packet transformation no matter how many layers the signal is decomposed, the series obtained in every channel will always keep the same length as that of the original signal. According to this speciality, a new denoising method based on module maximum value in wavelet packet transformation is proposed and realized. The idea of the denoising algorithm comes from the denoising algorithm based on module maximum value in wavelet transformation. But the essence of wavelet packet decomposition is to make a more decomposition to the decomposing result of wavelet′s, therefore the more high frequency noise is be separated in the posterior channel by wavelet packet and the more high frequency noise can be erased. So the denoising method based on wavelet packet transformation has the better effect than the one based on wavelet transformation. The detailed step of this denoising algorithm is given and two denoising examples using this method demonstrate that, compared with denoising result of the wavelet′s, the wavelet packet has the more excellent denoising effect.

关 键 词：信号分析 卷积型小波包变换 信号消噪算法 随机噪声 

分 类 号：TN911.6[电子电信—通信与信息系统]