Hybrid Feature Extraction Technique for Electrocardiograph Arrhythmia Signal Classification

This proposed work suggests, a helpful method for classifying arrhythmia beats from an ECG database. Three stages make up the recommended method namely the initial preprocessing phase, hybrid feature extraction stage and the feature classifier. At first, ECG beat signal is acquired from the physiobank ATM and are pre-processed in order to get them ready for feature extraction. A hybrid feature extractor has been employed for effective feature extraction. Two processes comprise the hybrid feature extraction process: (i) Haar wavelet-based feature extraction ii) feature extraction on the basis of tri-spectrum. Once the beat signal has been classified as normal or abnormal, the feature Feed Forward Neural Network (FFNN) classifier is employed. Studies on beat categorization are done using the MIT-BIH Arrhythmia ECG signal database. A classification accuracy of about 73 % is achieved using this beat classification method.