Classification of Arrhythmia using Time-domain Features and Support Vector Machine

Abstract

Cardiac arrhythmia is a heart condition where the heart does not beat in a regular way. This is one of those diseases which are easy to diagnose. A doctor can detect arrhythmia by just looking at the Electrocardiogram (ECG) of the patient because it has many visual clues, which a doctor is trained to identify. All these visual clues are the time-domain feature. Hence, in this paper, an algorithm is presented which uses only time-domain features to classify between normal sinus rhythm and arrhythmia using Support Vector Machine (SVM). The paper also compares the classification results when the frequency domain features are used along with the time-domain features. The frequency-domain features increase the computational complexity of the algorithm and make it harder to create a portable and reliable hardware device for the realtime detection and classification of arrhythmia. The proposed algorithm can be incorporated in a portable, lightweight and robust device which can detect arrhythmia in real-time. The accuracy of the algorithm is 99.36% on MIT-BIH arrhythmia database, which in comparison to other algorithm is an improvement.