基于分段香农能量与结构特征提取的心电R波检测方法  

Electrocardiogram R-Wave Detection Method Based on Shannon Energy Segmentation and Structure Feature Extraction

作　　者：张靖峰 谢志军[1] ZHANG Jingfeng;XIE Zhijun(Faculty of Information Science and Engineering,Ningbo University,Ningbo Zhejiang 315211,China)

机构地区：[1]宁波大学信息科学与工程学院,浙江宁波315211

出　　处：《传感技术学报》2023年第1期99-107,共9页Chinese Journal of Sensors and Actuators

基　　金：国家自然科学基金重点项目(U20A20121);浙江省自然基金项目(LY21F020006);宁波市“科技创新2025”重大专项(2019B10125,2019B10028,20201ZDYF020077);衢州市科技计划项目(2020K06);宁波市自然科学基金项目(2019A610088);宁波市公益项目(202002N3109)。

摘　　要：R波检测是采用心电图(electrocardiogram, ECG)评估心率变异性与检测其他特征波的基础。针对便携式采集环境中受运动伪迹与噪声影响的ECG记录存在R波检测准确度低的问题,提出了一种基于分段香农能量与结构特征提取的R波检测方法。首先采用平稳小波变换(Stationary Wavelet Transforms, SWT)提取原始ECG信号中的R波主要成分,然后通过分段香农能量增强R波特征信息,利用Hilbert变换的正交性质提取候选点,再采用结构特征分析校验候选点,最后根据心率标准优化结果,实现检测R波。将该方法应用4个典型ECG数据库,并与4种应用广泛的检测方法进行比较。在3个受噪声影响的ECG数据库检测实验中,该方法在R波检出个数方面分别提高了1.7%、20.17%与30.18%以上,F1分数指标上分别至少提高了1.3%、5.12%和4.7%。在临床ECG数据库检测实验中,所提方法的灵敏度、阳性准确率与F1分数均达到了99%以上。R-wave detection in electrocardiogram is the basis for assessing heart rate variability and detecting other characteristic waves. Targeting at the low R-wave detection accuracy for ECG recordings caused by motion artifacts and noise in a portable acquisition environment, an R-wave detection method based on segmented Shannon energy and structure feature extraction is proposed. Firstly, the R-wave major components of the original ECG signal are extracted by using the stationary wavelet transform, and then the R-wave feature information is enhanced through Shannon energy segmentation. The candidate points are extracted according to the orthogonal nature of Hilbert transform, then the candidate points are calibrated by using structural feature analysis. Finally, the R-wave is detected according to the optimization result of heart rate criterion. The method is applied to 4 typical ECG databases and compared with 4 widely used detection methods. In the detection experiments of three noise-affected databases, the number of R-wave detected by using this method is increased by more than 1.7%,20.17%,and 30.18%,respectively, and the F1 score index is increased by at least 1.3%,5.12%,and 4.7%,respectively. In the detection experiments of clinical ECG databases, the sensitivity, positive accuracy and F1 score of the proposed method reach more than 99%.

关 键 词：心电信号 R波检测 平稳小波变换 分段香农能量 结构特征提取 

分 类 号：TN911.7[电子电信—通信与信息系统]