Artificial neural network based ECG feature extraction using wavelet transform

Agrawal, Priyanka and Arun, Vanya and Basu, Amlan (2024) Artificial neural network based ECG feature extraction using wavelet transform. In: Second International Conference on Emerging Wireless Technologies and Sciences-2024, 2024-10-06 - 2024-10-07. (In Press)

[thumbnail of Agrawal-etal-ICEWTS-2024-Artificial-neural-network-based-ECG-features] Text. Filename: Agrawal-etal-ICEWTS-2024-Artificial-neural-network-based-ECG-features.pdf
Accepted Author Manuscript
Restricted to Repository staff only until 1 January 2099.
Download (899kB) | Request a copy

Abstract

In this research, Automatic techniques to detect diseases have been developed be-cause of requirements of continuous attention to patient having heart diseases. This research deals with implementation of Artificial neural network methods for analyz-ing ECG (Electrocardiogram) signals with a focus on early and accurate detection. Feature extraction of ECG signal plays vital role in cardiovascular diseases. ECG signal is decomposed using wavelet transform and then feature extracted of decom-posed ECG signal are given as input to Neural Network. The wavelets used for de-composition are Daubechies and Symmetric. The selection of detail coefficient d4 had been done based on the following important parameters i.e. Energy, Frequency and Correlation. The overall of detection using db6 and sym11 were 96.65% and 84.37%. In this work, study of the classification of ECG signal has been done in de-tail by using computational methods effectively for early cardiovascular diagnosis. Coefficients of discrete wavelet transforms are used for analyzing ECG signals in conjunction with the Artificial Neural network (ANN). Three different types of ECG data have been used normal sinus rhythm, supra ventricular arrhythmia and atrial fibrillation. Decomposition and Classification of ECG signals using discrete wavelet transform and Artificial Neural Network have been successfully designed. The meth-od has been implemented on 18 subjects. The results show that proposed method is effective for classification of normal and cardiac arrhythmia with an overall accura-cy of 97.5%.

ORCID iDs

Agrawal, Priyanka, Arun, Vanya and Basu, Amlan ORCID logoORCID: https://orcid.org/0000-0002-0180-8090;
CORE (COnnecting REpositories)