Comparative Evaluation of Denoising Algorithms for Enhanced SCG Signal Processing during Dynamic Conditions

Seismocardiography (SCG) is an emerging noninvasive technique for monitoring cardiac activity, particularly in wearable systems. However, motion artifacts significantly degrade SCG signal quality, especially during exercise, limiting its reliability in real-world applications. This study presents a comparative evaluation of denoising algorithms to enhance SCG-based heart rate estimation in dynamic conditions. We investigate seven denoising methods-Empirical Mode Decomposition (EMD), Ensemble EMD (EEMD), Complementary EEMD (CEEMD), Variational Mode Decomposition (VMD), Savitzky-Golay filtering, moving average filtering, and wavelet decomposition-alongside four heart rate estimation approaches based on peak detection, enveloping, and the Teager-Kaiser energy operator. SCG data were collected from 20 participants using a custom wearable patch during rest and stepping exercise. Results show that VMD and Savitzky-Golay filtering, when combined with enveloping, achieved the lowest mean absolute percentage error (MAPE) and root mean squared error (RMSE), reducing heart rate estimation error by up to 38% compared to unprocessed signals during exercise. These findings highlight the importance of signal processing techniques in improving SCG-based monitoring in real-world ambulatory settings, including during physical activity.