Neural Network Based Method for Automatic ECG Arrhythmias Classification

Automatic classification of electrocardiogram (ECG) arrhythmias is essential to timely and early diagnosis of conditions of the heart. In this paper, a new method for ECG arrhythmias classification using Wavelet Transform (WT) and neural networks (NN) is proposed. Here, we have used a discrete Wavelet Transform (DWT) for processing ECG recordings, and extracting some time-frequency features. In addition, we have combined the features extracted by DWT with ECG morphology and heartbeat interval features, to obtain our final set of features to be used for training a Multi-Layer Perceptron (MLP) neural network. The MLP Neural Network performs the classification task. In recent years, many algorithms have been proposed and discussed for arrhythmias detection. the results reported in them, have generally been limited to relatively small set of data patterns. In this paper 26 recordings of the MIT-BIH arrhythmias database have been used for training and testing our neural network based classifier. The simulation results of best structure show that the classification accuracy of the proposed method is 94.72% over 360 patterns using 26 files including normal and five arrhythmias.