ECG signal classification and parameter estimation using multiwavelet transform
Balambigai Subramanian*
Department of Electronics and Communication Engineering, Kongu Engineering College, Tamil Nadu, India
*Corresponding Author:
Balambigai Subramanian
Department of Electronics and Communication Engineering
Kongu Engineering College, India
Accepted on December 12, 2016
Abstract
The electrocardiogram (ECG) shows the plot of the bio-potential generated by the activity of the heart and is used by physicians to predict and treat various cardio vascular diseases. The QRS detection is a very important step in ECG signal processing. The main parameters concerned with QRS detection are sensitivity, accuracy, positive prediction and detection error. The methods used to detect QRS complex in ECG signals are Pan Tompkins algorithm, derivative based QRS detection and wavelet transform based QRS detection. In this paper performance comparison of wavelet transform based QRS detection with Pan Tompkins algorithm and derivative based QRS detection is done based on the characteristics of sensitivity, positive prediction and detection error. The accuracy of the proposed methodology is 93.35% and the specificity is 90%.
Keywords
Baseline wander, Electrocardiogram, Feature extraction, Multiwavelet transform, QRS complex.
Introduction
Electrocardiography (ECG) is a graphical representation of the electrical activity of the heart over a period of time which is recorded by the electrodes connected to the body either using three leads or twelve leads attached to the surface of the skin. [1]. The human circulatory system consists of heart, lungs and arteries along with veins that are helpful for the oxygen, nutrients and waste through the body. One cardiac cycle of ECG signal consists of the P wave QRS complex along with T waves is shown in Figure 1. P wave represents depolarization and the QRS represents ventricular depolarization. T wave represents rapid repolarization of the ventricles. During the measurement of ECG, there is a chance of various artefacts being recorded along with the signal such as power-line interference, ECG baseline wander, etc. [2]. These artefacts may cause a wrong diagnosis or misinterpretation of the vital cardiac parameters. The QRS complex is the most significant part in the ECG because of its high amplitude compared to P and T waves and uses in R peak detection to find the heart rate of a person. The digital signal processing techniques to be needed to filter out unwanted signal noise such as baseline drift motion artifact and powerline noise to provide accurate information from the input signals received from the electrodes [3]. Meyer et al. [5] proposed combining Pan – Tompkins and wavelet algorithms algorithms in automatic detection of QRS complexes in ECG signals. QRS complex and specifically Rpeak detection is the crucial step in every automatic electrocardiogram analysis. The Pan Tompkins algorithm does not adapt to the variations in the QRS complex structure and does not accurately provide useful information for the timevarying morphology of the QRS complex.
Figure 1: Normal ECG waveform [4].
Wavelet transforms are efficient for non-stationary signals that can detect the QRS complex morphology and its properties which changes with time and also to the noise in the signal. Arzeno et al. [6] proposed analysis of first derivative based QRS detection algorithms. In this, the algorithms based on the differentiated ECG are computationally efficient and ideal for real-time analysis of large datasets. This algorithm analyze traditional first – derivative based square function and Hilbert transform based methods for QRS detection and their modifications with improved detection thresholds. On a standard dataset, the Hamilton – Tompkins algorithm had the highest detection accuracy and largest timing error. The modified Hamilton Tompkins algorithm as well as the Hilbert transform based algorithms had comparable and slightly lower accuracy. The derivative based algorithm may miss many QRS complexes and is not robust to noisy signals and also its flexible in analyzing the time varying structure of ECG data.
Bagde and Raikwar [lt;a data-cke-saved-href='
剩余内容已隐藏,支付完成后下载完整资料
多小波变换的心电信号分类与参数估计
印度泰米尔纳德邦孔库工程学院电子与通信工程系
*通讯作者:Balambigai Subramanian
印度孔库工程学院 电子与通信工程系
2016 年 12月12日接受
抽象
心电图(ECG)显示由心脏活动产生的生物电位图,并由医生用于预测和治疗各种心血管疾病。QRS检测是ECG信号处理中非常重要的一步。与QRS检测有关的主要参数是灵敏度,准确度,正预测和检测误差。用于检测ECG信号中的QRS复合波的方法是Pan Tompkins算法,基于导数的QRS检测和基于小波变换的QRS检测。本文基于灵敏度,正预测和检测误差的特点,对基于小波变换的QRS检测与Pan Tompkins算法和基于导数的QRS检测进行了性能比较。拟议方法的准确性为93.35%,特异性为90%。
关键词
基线漂移,心电图,特征提取,多小波变换,QRS波群。
介绍
心电图(ECG)是心脏在一段时间内的电活动的图形表示,其通过使用连接到皮肤表面的三根引线或十二根引线连接到身体的电极记录。[ 1 ]。人体循环系统由心脏,肺和动脉以及静脉组成,这些静脉有助于通过身体吸收氧气,营养和废物。ECG信号的一个心动周期由P波QRS复合波和T波组成,如图1所示。P波表示去极化,QRS表示心室去极化。T波代表心室的快速复极化。在心电图的测量过程中,有可能记录各种伪像,如电源线干扰,心电图基线漂移等信号[ 2]]。这些假象可能导致对重要心脏参数的错误诊断或误解。QRS波群是心电图中最重要的部分,因为它与P波和T波相比具有高振幅,并且在R峰值检测中用于寻找人的心率。需要数字信号处理技术来滤除不需要的信号噪声,例如基线漂移运动伪影和电力线噪声,以便从电极接收的输入信号中提供准确的信息[ 3 ]。梅耶等人。[ 5提出结合Pan - Tompkins和小波算法算法自动检测心电信号中的QRS波群。QRS复合波,特别是Rpeak检测是每个自动心电图分析的关键步骤。Pan Tompkins算法不适应QRS复合结构的变化,并且不能准确地为QRS复合波的时变形态提供有用的信息。
图1:正常心电图波形[ 4 ]。
小波变换对于非平稳信号是有效的,该非平稳信号可以检测QRS复合波形态及其随时间变化的特性以及信号中的噪声。Arzeno等。[ 6提出了基于一阶导数的QRS检测算法的分析。在此,基于差异化ECG的算法在计算上是高效的,并且对于大数据集的实时分析是理想的。该算法分析了传统的基于一阶导数的平方函数和基于希尔伯特变换的QRS检测方法及其改进检测阈值的修改方法。在标准数据集上,Hamilton - Tompkins算法具有最高的检测精度和最大的定时误差。改进的Hamilton Tompkins算法以及基于希尔伯特变换的算法具有可比较且略低的精度。基于导数的算法可能遗漏许多QRS复合波,并且对噪声信号不具有鲁棒性,并且其在分析ECG数据的时变结构方面也是灵活的。
Bagde和Raikwar [ 7 ]提出了使用Matlab基于经验模式分解检测ECG波形的QRS波群。经验模式分解(EMD)方法依赖于完全数据驱动的机制,其不需要任何先验已知的基础。但是,它有相当大的延迟。Nazmy等人引用了各种心电图分类方法。[ 8 ]。
现有的QRS检测方法
QRS复合检测有助于找到与心跳相关的R峰。许多算法用于检测QRS形态,以帮助诊断任务,这些任务可能基于QRS检测期间获得的值的变化而出现[ 9]。在QRS检测期间要考虑的主要参数是准确性,预测性,一致性和硬件需求。增强方法是可用于增强ECG波形的特征和重要信息的技术[ 10 ]。
潘汤普金斯算法
Pan和Tompkins引入的QRS检测算法是最广泛使用且经常引用的用于从心电图中提取QRS复合波的算法。Pan Tompkins算法的框图步骤如图2所示。
图2: Pan Tompkins算法的框图。
所遵循的方法是ECG通过低通和高通滤波器以从信号中去除噪声。然后,滤波后的信号通过导数,平方和窗口积分阶段。最后,应用阈值技术并检测Q,R,S峰。Pan Tompkins算法的主要缺点是它们不适应QRS复合结构的变化,并且带通滤波器技术的缺点是带宽的前缀是它不能准确地为时变形态提供有用的信息。 QRS综合体 因此,需要具有一种通用的自适应方法,该方法能够有效地解决所述问题的局限性,这些问题覆盖了QRS形态的光谱和时间变化。
基于导数的QRS检测器
初始滤波器级通常由所有QRS检测算法使用,因为QRS复合波的典型频率分量在约5Hz至25Hz的范围内。这在实际QRS检测之前完成,以抑制ECG信号中的剩余特征,即P,T波,噪声和基线漂移。通过使用低通滤波器抑制噪声和基线漂移,而使用高通滤波器抑制P和T波等其他分量。基于导数的QRS检测器的框图如图3所示。因此,低通滤波器和高通滤波器组合在一起导致应用具有5Hz和25Hz的截止频率的带通滤波器用于QRS检测。
图3:基于导数的QRS检测器的框图。
对于许多算法,分别执行高通和低通滤波。一些算法仅使用高通滤波器。然后使用经滤波的信号使用与阈值的比较来检测QRS复合波。因此,具有大斜率的QRS复合特征被用于其检测。其缺点是基于导数的算法确实错过了许多QRS复合波,其由假阴性值的数量表示。基于导数的QRS检测器检测器对噪声信号不鲁棒,并且在分析ECG数据的时变结构时不灵活。因此,基于导数的QRS检测器的性能降低了。
特征选择
ECG波中R峰值的检测是特征提取的第一步。与其他导线相比,来自改良导联II(MLII)导线的信号中的R峰值在所有波形中具有最大幅度。所考虑的第二个特征是正常的QRS波群,其表明电脉冲通过右束和左束分支正常地从其束流到浦肯野网络,并且已经发生了右心室和左心室的正常去极化。QRS波群的重要性在于QRS波群的能量介于3 Hz和40 Hz之间。表1和2给出了正常ECG的幅度和定时信息。
波 |
振幅 |
---|---|
P波 |
0.25 mV |
R波 |
1.6 mV |
Q波 |
R波的25% |
T波 |
0.1至0.5 mV |
表1.正常ECG的幅度值。
间隔 |
持续时间 |
---|---|
PR波 |
0.12-0.2秒 |
QT波 |
0.35-0.44秒 |
ST波 |
0.05-0.15秒 |
P波间隔 |
0.11秒 |
表2.正常ECG的定时信息。
信号分类系统:自适应神经模糊推理系统基于Sugeno型模糊推理系统,由五层节点组成,其中第一层和第四层由自适应节点组成,第二层,第三层和第五层由固定节点组成。自适应节点与它们各自的参数相关联,随着每次后续迭代
剩余内容已隐藏,支付完成后下载完整资料
资料编号:[18793],资料为PDF文档或Word文档,PDF文档可免费转换为Word
课题毕业论文、外文翻译、任务书、文献综述、开题报告、程序设计、图纸设计等资料可联系客服协助查找。