心电信号特征提取及心律失常分类算法研究

发布时间：2018-03-05 16:34

本文选题：心电分析　切入点：经验模式分解　出处：《天津工业大学》2017年硕士论文　论文类型：学位论文

【摘要】：近年来,随着人们物质生活水平的提高,心血管疾病的发病率和死亡率逐年增加,并呈现出明显的年轻化趋势。而心血管疾病患者早期症状往往都伴随着心律失常现象,因此准确而及时的检测出患者心律失常类型,对预防心血管疾病具有及其重要的意义。心律失常分类技术是心电信号自动分析领域的重点研究内容,但由于其心电信号个体差异和易受噪声干扰的特点,要实现准确的特征提取及分类仍然存在一些难题。基于此,本课题针对心电信号特征提取及心律失常的分类进行了研究,本文的主要工作内容如下:1.心电信号预处理。本文分别针对心电信号中常见的低频基线漂移噪声及高频干扰噪声设计了中值滤波器及小波软阈值滤波器,并通过实验仿真验证,选取合适的窗口长度及小波基,较好的保留了原始信号的波形特点。2.心电信号特征提取。为了更加准确全面的表征心电信号的本质特征,本文提出了时域特征和变换域非线性特征相结合的方法。时域上通过经验模式分解和差分阈值相结合的方法提取了 QRS波群特征点,选取了 RR间期,心率变异性及QRS波群时限长度作为时域特征向量。利用经验模式分解及近似熵相结合的方法,通过对其前六个本证模态函数近似熵的计算,得到了心电信号变换域非线性特征。将两组特征融合作为分类特征向量集,为后续心电信号准确分类奠定基础。3.心律失常分类。综合比较几种常见分类器性能,选取对小样本非线性分类问题具有绝对优势的支持向量机分类模型对正常心电及四种常见心律失常信号进行分类处理。并针对标准粒子群参数优化算法在实际应用中易陷入局部最优的缺点,提出了改进的粒子群参数寻优算法。并综合利用改进的粒子群优化算法寻求最优参数,提高了分类的可靠性。综上所述,本文利用时域特征及其变换域非线性特征融合的特征向量集来表征心电信号,并利用改进的粒子群优化的支持向量机实现了常见心律失常信号的分类。通过MIT-BIH心律失常数据库进行仿真验证表明,本文算法能够实现心电节拍的准确分类,对心律失常诊断分析具有一定的现实意义,可用于心电分析辅助诊断。
[Abstract]:In recent years, with the improvement of people's material standard of living, the morbidity and mortality of cardiovascular diseases have been increasing year by year, and the trend of younger age is obvious. The early symptoms of patients with cardiovascular diseases are often accompanied by arrhythmia. Therefore, accurate and timely detection of patients' arrhythmia types is of great significance in preventing cardiovascular disease. Arrhythmia classification technology is an important research content in the field of ECG automatic analysis. However, due to the individual difference of ECG signal and its characteristics of being susceptible to noise interference, there are still some difficulties in the realization of accurate feature extraction and classification. Based on this, this paper studies ECG feature extraction and arrhythmia classification. The main work of this paper is as follows: 1. ECG signal preprocessing. In this paper, median filter and wavelet soft threshold filter are designed for low frequency baseline drift noise and high frequency interference noise respectively. Selecting the appropriate window length and wavelet base, the waveform characteristics of the original signal are better preserved. 2. ECG signal feature extraction. In order to more accurately and comprehensively characterize the essential characteristics of ECG signal, In this paper, a method of combining time domain features with transform domain nonlinear features is proposed. In time domain, QRS wave group feature points are extracted by empirical mode decomposition and difference threshold method, RR interval is selected. Heart rate variability (HRV) and the time limit of QRS wave group are used as time domain Eigenvectors. Using the method of empirical mode decomposition and approximate entropy, the approximate entropy of the first six intrinsic modal functions is calculated. The nonlinear features of ECG signal transform domain are obtained. The fusion of two groups of features as the classification feature vector set lays a foundation for accurate classification of ECG signals. 3. Arrhythmia classification. The performance of several common classifiers is compared synthetically. The support vector machine (SVM) classification model, which is superior to the small sample nonlinear classification problem, is selected to classify normal ECG and four kinds of common arrhythmia signals. The standard particle swarm optimization algorithm is applied in practice. It is easy to fall into local optimum in use, An improved particle swarm optimization algorithm is proposed, and the improved particle swarm optimization algorithm is used to find the optimal parameters, which improves the reliability of the classification. In this paper, the feature vector set of time domain feature and its transform domain nonlinear feature fusion is used to represent ECG signal. The classification of common arrhythmia signals is realized by using improved particle swarm optimization support vector machine. The simulation results of MIT-BIH arrhythmia database show that the proposed algorithm can achieve accurate classification of ECG beats. The diagnosis and analysis of arrhythmias have certain practical significance and can be used in ECG analysis to assist diagnosis.
【学位授予单位】：天津工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：R541.7;TN911.7

【参考文献】