Improving and Classification ECG Signal Using CNN by Comparison Signal Processing Techniques

Abstract

Detecting health issues is critical for saving patients' lives, especially when they are far from medical professionals, highlighting the importance of building highly accurate classification models. In this paper, a Convolutional Neural Network (CNN) model was designed and tested for analyzing and classifying electrocardiogram (ECG) signals. The main objective of this paper is to classify five types of heartbeats using a CNN model trained on the MIT-BIH Arrhythmia Database. Preprocessing techniques such as Butterworth band-pass filtering and Discrete Wavelet Transform (DWT) were applied to enhance the signals. The comparison between the two methods showed that the wavelet transform was more effective in denoising and preserving the essential signal components. A classification accuracy of 89.82%, recall of 69.61%, and an F1 score of 77.66% were achieved using wavelet-transformed pre-processed data, compared to an accuracy of 83.30% for the band-pass filtered data.