Hybrid EMD and RPF for Robust ECG Denoising and State Estimation

Electrocardiogram (ECG) signals are often corrupted by different noise sources such as baseline wander, power-lineinterference, and motion artifacts, which reduce the reliability of cardiac analysis and feature extraction. To overcome these issues, an enhanced ECG denoising approach based on Empirical Mode Decomposition (EMD) combined with a Regularized Particle Filter (RPF) is employed.First, the ECG data from the MIT-BIH database undergoes signal conditioning, including high pass, notch, and bandpass filtering to suppressdrift, interference, and unwanted noise while preserving the useful physiological components. The conditioned signal is then decomposed usingEMD, wherenoise-dominant IMFs are removed and selected components are reconstructed to obtain aleanerinputforfurtherestimation. To improve estimation stability, a Regularized Particle Filter with residual resampling is applied, whichreducesparticle degeneracy andmaintains particle diversity during sequential processing. Finally, Savitzky Golay smoothing is used to retain ECG morphology whileminimizing remaining fluctuations. The performance is evaluated using MSE, SNR, and PRD along with time and frequency domain analyses, effective noise reduction, improvedsignal reconstruction, and reliable preservation of key cardiac features such as R-peaks and RR intervals. KEY WORDS:ECG Denoising, Empirical Mode Decomposition,Regularized Particle Filter, State Estimation, R-Peak Detection, RR Interval Analysis, Biomedical SignalProcessing, Noise Removal.