Abstract

Info

Title:: Feature extraction and classification of arrhythmia using ensemble local mean decomposition method

Author(s):: CHEN Min; WANG Raofen; School of Electronic and Electrical Engineering, Shanghai University of Engineering Science, Shanghai 201620, China

Keywords:: ; Keywords: electrocardiogram signal; ensemble local mean decomposition; feature extraction; arrhythmia classification; support vector machine

Abstract:: Abstract: Aiming at feature extraction in electrocardiogram (ECG) automatic classification technology, a new method of feature extraction, ensemble local mean decomposition, is proposed. Firstly, different Gaussian white noises are added to ECG signals. Then, several product function (PF) components are obtained by local mean decomposition, and the average value after multiple decompositions is calculated as the final PF components. This method can solve the mode aliasing problem of local mean decomposition by multiple additions of noise and component averaging. The top 4 features of PF components are selected for feature calculation. The obtained feature vector matrix is used in support vector machine to classify normal ECG signals and 4 common kinds of arrhythmia ECG signals. Finally, the verification with MIT-BIH arrhythmia database shows that the total classification accuracy of the proposed method reaches 99.61%, higher than conventional methods, which confirms the effectiveness of ensemble local mean decomposition method.