Modeling of ECG Signal and Validation by Neural Networks

Abstract

ECG signal classification is widely used for diagnosing many cardiac diseases, which is the main cause of mortality in developed countries. Since most of the clinically useful information in the ECG signal is found in the intervals and amplitudes. The development of accurate and robust methods for automatic ECG signal classification is a subject of major importance. Modeling techniques like Least Square Estimation (LSQ) and Autoregressive (AR) modeling have been performed on the ECG signal. The model coefficients extracted using autoregressive modeling technique was found to be resourceful, so it has been taken for further validation. The ECG data is taken from standard MIT-BIH Arrhythmia database. AR coefficients obtained from the AR modeling are fed to the back-propagation neural network which classifies the ECG signal. In order to train the modeling coefficients with the back-propagation neural network the architecture implemented with 2 input neurons, 2 hidden neurons and 2 output neurons. In this work all neurons uses sigmoid activation function.