论文部分内容阅读
心律失常是一种极其常见的心电活动异常症状,基于心电图(ECG)的心拍分类对心律失常的临床诊断具有十分重要的意义。本文提出一种基于流形学习的特征提取方法——近邻保持嵌入(NPE)算法,实现心律失常心拍的自动分类。分类系统利用NPE算法获取高维心电节拍信号的低维流形结构特征,然后将特征向量输入支持向量机(SVM)分类器进行心拍的分类诊断。实验基于MIT-BIH心律失常数据库提供的ECG数据,对14种类型的心律失常心拍进行分类,总体分类准确率高达98.51%。实验结果表明,所提方法是一种有效的心律失常心拍分类方法。
Arrhythmia is a very common abnormal ECG symptoms, ECG-based heart beat classification is very important clinical diagnosis of arrhythmia. In this paper, we propose a feature extraction method based on manifold learning, namely, neighbor-preserving embedding (NPE) algorithm, which can automatically classify cardiac beat of arrhythmia. The classification system uses the NPE algorithm to obtain the low-dimensional manifold features of the high-dimensional ECG beat signals, and then inputs the feature vectors into the SVM classifier to perform heart beat classification and diagnosis. Based on the ECG data provided by the MIT-BIH arrhythmia database, the experiment categorized 14 types of cardiac arrhythmias with overall classification accuracy of 98.51%. The experimental results show that the proposed method is an effective cardiac arrhythmia classification method.