ECG classification using wavelet transform and wavelet network classifier

Abstract

An electrocardiogram (ECG) signal is a highly used examination in the field of cardiology; these pathologies are generally reflected by disorders of the electrical activity of the heart. In this paper, we have addressed the problem of automatic recognition of heartbeats through the development and implementation of a method combining wavelet transform with neural networks. This method consists of denoising, extraction, and classification models for robust automated ECG analysis. For the classification module, a hybrid network combining neural networks and wavelets has been proposed, implemented and evaluated for identification of the ECG classes. This technique is based on use of wavelet functions as activation function in neural networks, which allowed the wavelet network to have a better adaptability and flexibility during the learning process given the parameters of translation and expansion of the functions of wavelets. Indeed, the evaluation of the results obtained by the implemented wavelet network is satisfactory with respect to other neural networks in terms of the rate of classification of the heartbeats. The association of neural networks with the wavelet functions made it possible to extract the strengths of the two techniques (the learning capacity of the neural models and the multiresolution analysis of the wavelets). The results obtained showed that the proposed method can be considered as an effective method for the classification of cardiac arrhythmias with a very acceptable accuracy of more than 98.78%.