Adaptive Self Regulating Denoising For ECG Signal Using SVM – Based Shrinkage Function Prediction

Electrocardiogram (ECG) signals are often affected by various forms of noise that can distort critical diagnostic features. Traditional wavelet-based denoising techniques rely on fixed thresholding strategies and shrinkage functions, which may not perform optimally across varying signal conditions. This paper presents an adaptive self-regulating denoising framework that dynamically selects the most suitable shrinkage function using a Support Vector Machine (SVM) classifier. The approach leverages time-domain and frequency-domain features extracted from noisy ECG segments to predict the ideal shrinkage function such as soft, hard, hyper, or negative garrote tailored to each signal window. This adaptive strategy improves denoising performance without manual intervention. Experimental results demonstrate the effectiveness of the proposed model in enhancing signal fidelity while preserving critical waveform characteristics. The self-regulating mechanism ensures robustness against diverse noise types, offering potential for real-time ECG monitoring systems.