Classification of Cardiac Diseases from ECG Signals Using Time-Varying Feature Extraction Techniques

Tan, Bing Jiat (2026) Classification of Cardiac Diseases from ECG Signals Using Time-Varying Feature Extraction Techniques. Final Year Project (Bachelor), Tunku Abdul Rahman University of Management and Technology.

Text Tan Bing Jiat_Thesis.pdf Restricted to Registered users only Download (1MB)

Abstract

The project focuses on developing an automated system to classify cardiac diseases from electrocardiogram (ECG) signals using time-varying feature extraction techniques and deep learning models. Traditional manual interpretation of ECG signals is often time-consuming, prone to human error, and challenging due to the non-stationary nature of the signals. In addition, ECG data in one-dimensional form makes it difficult to capture important frequency-related information for accurate diagnosis. Therefore, this study aims to improve classification performance by transforming ECG signals into more informative representations and applying an enhanced deep learning approach. The methodology involves a hybrid feature extraction process, where Variational Mode Decomposition (VMD) is first used to decompose the signals, followed by Short-Time Fourier Transform (STFT) to generate time–frequency spectrograms. These spectrograms are then used as input to a Convolutional Neural Network (CNN) integrated with an attention module, allowing the model to focus on the most relevant features. The system is trained and evaluated using the MIT-BIH Arrhythmia dataset, with performance assessed using accuracy, precision, recall, F1-score, and confusion matrix. he results show that the proposed CNN with attention module achieves an accuracy of approximately 98%, outperforming the preliminary VGG16 model, which achieved around 97%. Overall, the proposed approach provides a more effective and reliable solution for cardiac disease classification