Title: Classification of cardiac arrhythmias based on morphological and rhythmic features

Authors: R. Shantha Selva Kumari; J. Ganga Devi

Addresses: Department of Electronics and Communication Engineering, Mepco Schlenk Engineering College, Sivakasi, Tamil Nadu, India ' Department of Electronics and Communication Engineering, Kamaraj College of Engineering and Technology, Virudhunagar, Tamil Nadu, India

Abstract: Cardiac arrhythmias stand a great admonish for human beings nowadays. The proposed work intends to classify four commonly occurring arrhythmia classes along with normal class. For each beat of 300 samples, both morphological and rhythmic features are determined. A total of 129 morphological features are formed by 114 wavelets coefficients and 15 independent components having 300 coefficients of basis functions obtained by using ICA. PCA is applied on the morphological features to derive the best 11 principal components and to this, four rhythmic features are combined to have a final 15 feature coefficients. SVM classifier gets trained using the 15 features of 30% beats of every class in the total number of beats. The remaining 70% of beats are used for evaluating the individual class performance. Finally the SVM classifier with only 15 features is able to produce the overall accuracy of 99.29% for a total 82,978 beats.

Keywords: PCA; principal component analysis; ICA; independent component analysis; wavelets; arrhythmia classification; cardiac arrhythmias; morphological features; rhythmic features; heartbeats; SVM; support vector machines; heart rate.

DOI: 10.1504/IJBET.2014.059669

International Journal of Biomedical Engineering and Technology, 2014 Vol.14 No.3, pp.192 - 208

Received: 14 Jun 2013
Accepted: 17 Jan 2014
Published online: 16 Oct 2014 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article