A COMPREHENSIVE QRS DETECTION METHOD BASED ON EXCLUSIVE MOTHER WAVELET AND ARTIFICIAL NEURAL NETWORK

Abstract

Detecting the QRS complex on an ECG signal leads to precious information about the signal under study. Different noises, arrhythmias, and diseases alter the shape and energy of the signal, making it harder to detect the QRS points. Several algorithms for QRS detection have been proposed and most of them merely focus on precision improvement, and therefore certain limitations have emerged with regard to deployment of these algorithms. As a result, while developing the new algorithm, not only efforts have been made to keep the precision at a high level, but also it has been tried to keep an eye on the generality of the algorithm, and to eliminate the end user limitations as much as possible. To this end, we have used an exclusive mother wavelet together with an artificial neural network to develop an algorithm which not only has superior precision, but also does not require changing the tuning parameters for each different signal. In other words, the algorithm extracts the required parameters automatically. In this method, first, an exclusive mother wavelet identical to the input signal is formed. Then, by using the mother wavelet, matrices containing sufficient data to be processed by the neural network are developed. Using these matrices, the existing QRSs will be detected with a sensitivity of 99.81% on MIT-BIH and 99.49% on physiozoo datasets.