改进CEEMD–MRSVD降噪方法及应用研究

Study on Improved CEEMD–MRSVD Denoising Method and Its Application

针对总体经验模态分解(Ensemble empirical mode decomposition,EEMD)阈值降噪处理方法难以取得理想效果问题,提出了改进互补集合经验模态分解–多分辨率奇异值分解(Complementary ensemble empirical mode decomposition–multi-resolution singular value decomposition,CEEMD–MRSVD)降噪方法。含噪信号经CEEMD处理,克服了EEMD时效性差及EMD模式混叠缺陷,然后利用信号和噪声的相关特性对分解得到的本征模态分量进行信号主导和噪声主导分量区分,根据噪声强度不同,提出对噪声主导和信号主导的本征模态分量进行策略性优化的多分辨率奇异值处理方法,最后经Savitzky–Golay平滑滤波,剔除信号粗糙细节,重构达到降噪目的。试验通过仿真信号和超声回波信号降噪处理,结果表明,此方法不仅有效剔除了噪声干扰,而且减少了有用细节流失。

In view of the problem that ensemble empirical mode decomposition(EEMD) threshold denoising method was difficult to achieve ideal results,an improved complementary ensemble empirical mode decomposition–multiresolution singular value decomposition(CEEMD–MRSVD) denoising method was proposed.Noisy signal was processed by CEEMD,which overcame the shortcomings of poor timeliness of EEMD and EMD mode mixing.Then,the signaldominated and noise-dominated components were distinguished by the correlation characteristics of signal and noise.According to the different noise intensity,a multi-resolution singular value location was proposed to strategically optimize the noise-dominated and signal-dominated components.Finally,the Savitzky–Golay smoothing filter was used to remove the rough details of the signal and the extracted signals were reconstructed to achieve the denoising purpose.Results showed that,this method can not only eliminated noise interference effectively,but also reduced the loss of useful details.