基于区间阈值和部分重构的MEMD去噪方法  

Novel Denoising Method Based on Interval Thresholding and Partial Reconstruction Using MEMD

作　　者：郝欢[1] 鞠少华[2] 王华力[3] 曾显华[4] 尹廷辉[3] 

机构地区：[1]解放军理工大学通信工程学院研究生4队,江苏南京210007 [2]中国人民解放军第89医院,山东潍坊261000 [3]解放军理工大学通信工程学院 [4]解放军理工大学通信工程学院研究生2队

出　　处：《军事通信技术》2016年第2期81-86,共6页Journal of Military Communications Technology

基　　金：国家自然科学基金资助项目(61271354)

摘　　要：经验模式分解(EMD)是一种自适应的非线性、非平稳信号分析方法,广泛用于非参数化信号去噪。由于存在边界效应和模式混叠现象,EMD去噪效果受到一定影响。为了提高去噪性能,文章提出一种基于区间软阈值和部分重构的多维经验模式分解(MEMD)去噪方法。该方法利用MEMD对于高斯白噪声的良好二进滤波特性,以固有模式函数(IMF)与输入信号概率密度函数(PDF)之间的相似度来选择最佳模式函数。根据IMF本身特点,采用区间软阈值去噪方法对选取的IMF分量进行去噪,最后结合部分重构实现信号的去噪。实验仿真和实测EEG信号处理结果表明,与小波变换和EMD-CIIT方法相比,文中方法对单通道信号中高斯噪声具有2dB^3dB的性能提升,同时还可以对多通道信号进行联合去噪,是一种有效的信号去噪新方法。Empirical Mode Decomposition（EMD） is a nonlinear and non-stationary signal a- nalysis method, which has been widely used for non-parametric signal denoising. However, the denoising performance of EMD is limited by the boundary effect and mode mixing. In order to en- hance the denoising performance, interval soft thresholding and partial reconstruction based Mul- tivariate Empirical Mode Decomposition（MEMD） denoising method was proposed in this paper. In the proposed method, the dyadic filterbank property of MEMD for white Gaussian noise was investigated, and then the similarity of IMF and the corresponding original signal was adopted to select relevant mode functions. According to the special characteristics of the signal modes resul- ting from MEMD, interval soft thresholding was adopted for the relevant modes and the signal reconstructed with partial reconstruction. Simulation and measured EEG signal processing results show that, compared with wavelet transform and EMD-CIIT methods, the proposed method is a novel adaptive denoising method, which not only achieves 2 dB-3 dB performance increase for uni-channel signal denoising contaminated by Gaussian noise, but also can be used for multi-chan- nel signal joint denoising.

关 键 词：经验模式分解 信号去噪 多维经验模式分解 固有模式函数 

分 类 号：TN911.72[电子电信—通信与信息系统]