改进WOA-VMD算法的心电信号去噪-《中国医学物理学杂志》

文章信息/Info

作者:: 郁伟1; 李正权1; 2; 邢松3; 1.江南大学物联网工程学院，江苏无锡 214122； 2.江苏理工学院常州市5G+工业互联网融合应用重点实验室，江苏常州 213001； 3.加利福尼亚州立大学信息系统系，洛杉矶 90032

Author(s):: YU Wei1; LI Zhengquan1; 2; XING Song3; 1. School of Internet of Things Engineering, Jiangnan University, Wuxi 214122, China 2. Changzhou Key Laboratory of 5G+ Industrial Internet Fusion Application, Jiangsu University of Technology, Changzhou 213001, China 3. Information Systems Department, California State University, Los Angeles CA 90032, USA

Keywords:: Keywords: electrocardiogram signal denoising whale optimization algorithm variational modal decomposition wavelet threshold correlation coefficient

摘要:: 传统方法进行心电图（ECG）信号分解是基于QRS特征波经验性识别固有模式函数变量（IMF），但ECG信号和噪声信号之间存在频带混叠，导致去噪效果较差，针对此问题提出一种改进鲸鱼算法（IWOA）优化变分模态分解（VMD）算法参数，并和改进小波阈值相结合的方法。利用VMD基于完全非递归分解的特性，在鲸鱼算法中引入感知扰动机制，并用等螺距阿基米德螺旋曲线代替经典的对数螺旋曲线，对VMD中的模态个数K和惩罚参数α进行寻优；然后对ECG信号进行VMD分解，得到一系列IMF分量，通过相关系数判定噪声分量，对其进行改进小波阈值去噪；最后将各分量重构，得到去噪后的信号。将本文方法与单独使用小波阈值算法或VMD算法进行去噪对比实验，结果表明，本文方法可以有效去除ECG信号中的各种干扰，更好地保留ECG信号波形特征，具有潜在的临床指导意义。

Abstract:: Abstract: The traditional method for decomposing electrocardiogram (ECG) signals is based on empirically identifying intrinsic mode functions (IMF) using QRS feature waves. However, there is a frequency band aliasing between ECG and noise signals, which significantly affects the denoising performance. To solve this problem, an improved whale optimization algorithm (WOA) is proposed to optimize the parameters of the variational mode decomposition (VMD) algorithm, and combined it with an improved wavelet threshold method. For VMD is based on the completely non-recursive decomposition, a perception disturbance mechanism is introduced into WOA, and an equal pitch Archimedean spiral instead of the classic logarithmic spiral is used to optimize the number of modalities K and the penalty parameter α in VMD. Then, the ECG signal is decomposed using VMD to obtain a series of IMF components. The noise components are determined by correlation coefficient and removed using the improved wavelet threshold method. Finally, the various components are reconstructed to obtain the denoised signal. The comparison with wavelet threshold algorithm and VMD algorithm shows that the proposed method (WOA-VMD algorithm) can effectively remove various interferences in the ECG signal and better retain the waveform characteristics of the ECG signal, which is of potential significance in clinic.