Improved ECG Signal Denoising with Notch Filter for Enhanced Cardiac Disease Classification

The wide frequency range of electrocardiogram (ECG) signals-from 0.05 to 100 Hz—is often muddled by background noise, which makes definitive diagnosis difficult. To preprocess ECG signals, this research suggests an Improved Notch Filter (iDENOISE) that decreases noise at both high and low frequencies. The iDENOISE filter improves upon traditional notch filtering by actively altering the depth and breadth of the notch in response to the noise. In terms of efficiency, iDENOISE is pitted against old-school filters like Adaptive Median Filter, Gabor Filter, and Recursive Least Square Notch Filter. Based on research results utilising the MIT-BIH Arrhythmia database (50,000 samples; 80% training, 20% testing), the RBFNN classifier obtains the maximum accuracy (99.8%) with iDENOISE. A combination of iDENOISE and PSO-PNN further improves categorisation effectiveness and runs the program quickly. Measures like F1-score (97.2), recall (95.8), and precision (98.7) demonstrate outstanding results. The research emphasises the efficacy of iDENOISE in real-time ECG evaluation, which optimises noise suppression while maintaining important signal qualities for better cardiac diagnosis.