ECG Arrhythmia Classification using Daubechies Wavelet and Radial Basis Function Neural Network

ECG arrhythmia classification have been performed using radial basis function neural network and multilayered perceptron to classify the five types of ECG beats: Normal beat, Paced beat, Left bundle branch block beat, Right bundle branch block beat and premature ventricular contraction beat in this paper. MIT-BIH arrhythmia database was utilized for the extraction of 500 ECG beat which are arbitrarily extracted from 26 records. Each ECG beats were represented by 21 points from p1 to p21 which are known as features and these ECG beats from each record were classified according to types of beats. The classification of ECG arrhythmia has been followed by preprocessing; R-peak detection and ECG beat extraction. The simulation results obtained for the classification result of ECG beats with average accuracy of 99.84%, sensitivity of 99.60% positive predictivity of 99.60%, specificity of 99.90%, classification error rate of 0.16%. The overall accuracy of 98.8% and 99.6% was achieved using BPNN and RBFNN classifier respectively.