Categorizing Heartbeats by Independent Component Analysis and Support Vector Machines

We propose a method that utilizes independent component analysis (ICA) and support vector machines to classify electrocardiogram (ECG) beats. In this study, ICA is used to dig up underlying components from ECG signals. A classifier constructed by support vector machines follows to categorize the input ECG beats into one of eight beat types. The independent components are calculated from the training ECG beats and serve as the bases of the system. The features based on ICA and the RR time interval between consecutive ECG beats are employed as inputs to the classifier. In the study, 9800 ECG samples, including eight different ECG types, were selected from the MIT-BIH arrhythmia database for experiments. The experiments showed the accuracy attained to 98.7% under the condition that 20 independent components were used. The results show the potential of the proposed method in the computer-assisted diagnosis of heart disorders based on ECG signals.