一种采用小波变换和动态窗法的心电信号识别方法

A ECG Signals Recognition Method by Using Wavelet Transform and Dynamic Window

心电信号中的P波、QRS波群、T波和U波包含了心脏跳动的大量信息,是诊断心脏疾病的重要依据之一。本文提出了一种采用小波变换和动态窗法的心电信号特征提取方法,并应用于心律失常类心电信号的识别。首先采用小波变换对心电信号数据进行分解重构,滤除信号中部分低频信息和绝大部分高频噪声后,通过巴特沃斯低通滤波器对信号进行包络提取,并采用动态窗法得到P波、QRS波群、T波和U波的信息特征,最终根据信息特征得到心电信号特征。采用麻省理工学院心电信号数据库(MIT-BIH)进行心律失常心电信号识别实验。实验结果表明,本文所提取的特征在BP神经网络、K最近邻、支持向量机、C4.5决策树及随机森林方法等5类机器学习算法下平均识别率达到了95.7%,随机森林分类器中特征识别率最高,达到了98.4%。实验表明所提取的特征对心律失常心电信号具有较高的识别性能。

The P wave, QRS wave group, T wave and U wave in ECG signal contain most of the information of heart beat, which are one of the important basis for the diagnosis of heart disease. In this paper, a method of ECG signal noise reduction and feature extraction is proposed and applied to ECG signal recognition of arrhythmia. Firstly, the wavelet transform is used to decompose and reconstruct the data to filter out the signal in the part of low frequency information, and the vast majority of high-frequency noise, then Butterworth low-pass filter is used to extract the signal envelope. Finally, the dynamic window method is used to obtain the information characteristics of P wave, QRS wave group, T wave and U wave. The feature information of ECG signal were obtained according to the information characteristics. The ECG signal recognition experiment of arrhythmia was carried out using MIH-BIN ECG Signal Database. The experimental results showed that the average recognition rate of features extracted in this paper reaches 95.7% under five machine learning algorithms including BP neural network, K-nearest Neighbor, support vector machine, C4.5 decision tree and random forest method, and the feature recognition rate is the highest in random forest. It reached 98.4%. It shows that the features extracted in this paper have high recognition performance for ECG signal of arrhythmia.