Abstract—This paper presents the classification of the cardiac arrhythmia whose features are extracted by applying the autoregressive signal modeling. The different coefficients of the 4th to 9th-order AR models were tested with the support vector machine as one-against-the-rest approach for each arrhythmia in order to investigate their performance. The results lead to the proposed method where the first step is to apply APC-against-non-APC SVM on the extracted AR features. Secondly, the PVC-against-Normal SVM classifier is applied to the non-APC feature classified from the first step. The mean value of the best overall accuracy is 94.40% which is obtained when the 6th-order AR model is applied to modeling. The APC-against-non-APC classification alone provides the high accuracy of 97.11%.

Index Terms—Electrocardiography, support vector machine, classification, autoregressive estimation.

C. Thanawattano is with National Electronic and Computer Technology Center, Pathumthani, Thailand (e-mail: chusak.tha@nectec.or.th ).
T. Yingthawornsuk is with Department of Media Technology, King Mongkut’s University of Technology Thonburi – Bang Khunthian Campus, Bangkok, Thailand (e-mail: thaweesak.yin@kmutt.ac.th).

Cite: Chusak Thanawattano and Thaweesak Yingthawornsuk, "Cardiac Arrhythmia Classification using Beat-by-Beat Autoregressive Modeling," International Journal of Computer and Electrical Engineering vol. 4, no. 5, pp. 762-765, 2012.