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

Area of Science:

• Biomedical Engineering
• Cardiovascular Monitoring
• Signal Processing

Background:

• Seismocardiography (SCG) offers noninvasive cardiac monitoring, ideal for wearables.
• Motion artifacts in SCG signals, especially during exercise, compromise data reliability.
• Accurate heart rate estimation in dynamic conditions remains a challenge for SCG.

Purpose of the Study:

• To comparatively evaluate various denoising algorithms for SCG signals.
• To assess the impact of denoising on heart rate estimation accuracy during dynamic activities.
• To identify optimal signal processing techniques for robust SCG-based monitoring.

Main Methods:

• Investigated seven denoising methods: EMD, EEMD, CEEMD, VMD, Savitzky-Golay, moving average, and wavelet decomposition.
• Employed four heart rate estimation approaches: peak detection, enveloping, and Teager-Kaiser energy operator.
• Collected SCG data from 20 participants during rest and stepping exercise using a wearable patch.

Main Results:

• Variational Mode Decomposition (VMD) and Savitzky-Golay filtering, coupled with enveloping, yielded the best performance.
• These methods reduced heart rate estimation errors (MAPE and RMSE) by up to 38% during exercise compared to unprocessed signals.
• Demonstrated significant improvement in SCG signal quality and heart rate accuracy in dynamic conditions.

Conclusions:

• Advanced denoising techniques are crucial for reliable SCG-based heart rate monitoring in ambulatory settings.
• VMD and Savitzky-Golay filtering show promise for enhancing SCG applications during physical activity.
• Improved SCG signal processing supports continuous, accurate cardiovascular assessment in real-world scenarios.