Highly accurate ECG beat classification based on continuous wavelet transformation and multiple support vector machine classifiers

This paper presents a highly accurate ECG beat classification system. It uses continuous wavelet transformation combined with time domain morphology analysis to form three separate feature vectors from each beat. Each of these feature vectors are then used separately to train three different support vector machine (SVM) classifiers. During data classification each of the three classifiers independently classifies each beat; with the result of the multi classifier based classification system being decided by voting among the three independent classifiers. Using this method the multi classifier based system is able to reach an average accuracy of 99.72% in the classification of six types of beats. This accuracy is higher than the individual accuracy of any of the participating SVM classifiers as well as higher than previously presented ECG beat classification systems showing the effectiveness of the technique.