Beat classification for use in arrhythmia analysis

Abstract

An important step in the classification of arrhythmias is beat classification. The aim of this work is to classify the beats in the MIT-BIH Arrhythmia database and to investigate pertinent features for beat classification. A comparison is made of the classification performance using features obtained after three different processing tasks on the electrocardiogram. QRS detection, QRS onset/offset detection and P wave onset detection. Linear discriminants and quadratic discriminants were employed to model the classifier. The best classification performance of 82% was achieved by employing the QRS width as an input feature to the linear discriminant classifier