A Wavelet-Based Approach for Motion Artifact Reduction in Ambulatory Seismocardiography

Area of Science:

• Biomedical Engineering
• Wearable Health Technology
• Cardiovascular Monitoring

Background:

• Wearable seismocardiography (SCG) shows promise for cardiac health monitoring.
• Motion artifacts, particularly from walking, significantly degrade SCG signal quality.
• This degradation limits the clinical applicability of SCG for continuous monitoring.

Purpose of the Study:

• To develop and validate techniques for enhancing SCG signal quality in the presence of walking motion artifacts.
• To create a standalone SCG-based solution for reliable cardiovascular monitoring during ambulation.
• To assess the effectiveness of the developed algorithm on heart rate estimation accuracy.

Main Methods:

• Simulated ambulant SCG data by corrupting clean signals with real-walking vibrational noise.
• Employed signal decomposition using maximum overlap discrete wavelet transform (MODWT).
• Developed a novel algorithm combining MODWT, time-frequency masking, and nonnegative matrix factorization, utilizing vertical axis accelerometer data to reduce walking vibrations.

Main Results:

• The MODWT was identified as the most effective decomposition method for motion artifact reduction.
• The novel algorithm improved heart rate estimation accuracy (R-squared) from 0.1 to 0.8 at -15 dB signal-to-noise ratio (SNR).
• The method successfully reduced motion artifacts in SCG signals up to -19 dB SNR without requiring electrocardiography (ECG) assistance.

Conclusions:

• The developed algorithm effectively mitigates motion artifacts in SCG signals, enabling reliable cardiovascular monitoring during walking.
• This standalone SCG solution enhances applicability for daily life monitoring and as a clinical wearable alternative.
• Incorporating ECG can further improve performance in higher noise environments, extending the usable range of SCG.