基于非负矩阵分解和支持向量机的心电图分类

ECG Classification Based on Nonnegative Matrix Factorization and Support Vector Machine

提出一种心电信号分类方法,利用非负矩阵分解进行数据降维,运用支持向量机进行心电信号分类,以保留更多的原始数据信息,从而更有效地提取高维心电数据特征,提高分类准确度。通过对MIT-BIH数据库中4类常见心电信号进行分类实验,证明该方法的整体准确率达到99%。

In order to achieve better Electrocardiograph（ECG） characteristics from high-dimensional data and realize accurate automatic ECG classification, a novel method for ECG multi-classification is proposed. This method uses Nonnegative Matrix Factorization（NMF） for data dimension reduction and conducts multi-classification by Support Vector Machine（SVM）. In implementing the conversion of high dimension to low dimension, NMF retains the original information and supplies better eigenvectors, so it improves the classification accuracy. By testing four kinds of ECG from the MIT-BIH arrhythmia database, the total accuracy is up to 99%.