RHYTHM IDENTIFICATION AND CLASSIFICATION FOR ELECTOCARDIOGRAM SIGNALS USING FEATURE CLUSTER FRAMEWORK CLASSIFIER

It is necessary to explore the feature segments of electrical representation of heart signals i.e. ECG signals for the arrhythmia identification. A feature cluster extraction based classification is proposed in our research for detection and identification of rhythm class. After the feature cluster extraction of signal record it is compared with the set of rules framework for rhythm class assignment. The work has been validated and compare with relevant state of art methods and found improved performance with respect to sensitivity Se and positive predictivity Pp parameters for ventricular and supraventricular rhythm class. Proposed approach achieved sensitivity (Se) as 94.25% and positive predictivity (Pp) as 96.35% and sensitivity (Se) 86.5% and positive predictivity (Pp) 83.47% for ventricular rhythm and supraventricular rhythm respectively for dataset DS2. As a result we achieved significant improvement after comparison with other methods for the MIT-BIHA database.