Early Detection of Arrhythmia with Advanced Feature Extraction Techniques

Early detection of arrhythmia is critically important for cardiac patients. This is achieved by analyzing electrocardiogram (ECG) signals and extracting discriminative features for classification of various arrhythmia types. In this paper, we implement three distinct feature extraction algorithms: Fast Fourier Transform (FFT) for frequency-domain analysis, Autoregressive (AR) modeling for time-series parameterization, and Principal Component Analysis (PCA) for dimensionality reduction. The classification is performed using an Artificial Neural Network (ANN) with backpropagation training. Our experiments show that the system utilizing PCA features achieves the highest accuracy. The proposed methodology processes full 3-second intervals for both training and testing data. Compared to existing studies on similar datasets, our approach attains an accuracy of 92.7083%, significantly outperforming the reference accuracy of 84.4%. The results demonstrate that PCA-based feature extraction effectively enhances arrhythmia identification accuracy. This method can be adapted for various arrhythmia recognition and classification tasks. Furthermore, these features show potential for predicting cardiac diseases, enabling improved treatment plans and patient care strategies through proactive diagnosis.