Electrocardiogram (ECG) signal is an electrical manifestation of contractile activity of the heart. For analysis of ECGs, it is desirable to classify the obtained signal accordingly for suitable diagnosis. Many challenges have been identified by various researchers in processing, analyzing and classifying ECG signals. This paper proposes a multimodal decision learning (MDL) algorithm for classification and arrhythmia identification. The features are extracted using Integrated Peak Analyzer is used and Intensity Weighted Fire-Fly Optimization is employed for feature reduction process. In post-processing stage, proposed MDL algorithm is employed for ECG classification and label identification. Six classes of ECG functions indicating different functioning conditions like Normal Sinus Rhythm (NSR), Ventricular Tachycardia (VT), First Degree AV Block (FDB), Supraventricular Tachycardia (SVTA), Atrial Fibrillation (AF), Ventricular Flutter (VF). The efficacy of the method is established by comparing it with the SVM based classifier. The metrics used for comparison include confusion matrix (CM), false rejection ratio (FRR), false acceptance ratio (FAR), global acceptance ratio (GAR), Kappa coefficient (KC), sensitivity, specificity and accuracy.
展开▼