A Wavelet-Based Approach for Motion Artifact Reduction in Ambulatory Seismocardiography
Executive Summary
This study introduces a novel algorithm combining Maximum Overlap Discrete Wavelet Transform (MODWT), time-frequency masking, and nonnegative matrix factorization (NMF) to mitigate motion artifacts in seismocardiography (SCG) signals during ambulatory scenarios. The algorithm demonstrated significant improvements in heart rate estimation accuracy, achieving an R-squared of 0.8 at -15 dB signal-to-noise ratio (SNR), and extended SCG's usability without reliance on electrocardiography (ECG). These advancements enhance SCG's potential for wearable cardiac monitoring in real-world and clinical settings.
Answer Machine Insights
Q: What is the primary advantage of the proposed algorithm?
The algorithm significantly reduces motion artifacts in SCG signals, enabling accurate heart rate estimation without relying on ECG.
Our method reduces motion artifacts in SCG signals up to a SNR of −19 dB without requiring any external assistance from electrocardiography (ECG).
Q: How does the algorithm perform under high noise conditions?
The algorithm improves heart rate estimation accuracy up to -19 dB SNR but struggles at full-strength noise levels (-27 dB SNR).
At −15 dB, we observed a noteworthy improvement in HR r2, elevating it from less than 0.1 to above 0.8. However, at −27 dB, the method only marginally increased the r2 from 0 to about 0.1.
Key Results
MODWT achieved the highest signal reconstruction accuracy with an R-squared of 0.42 at -20.9 dB SNR.
Heart rate estimation improved from an R-squared of 0.1 to 0.8 at -15 dB SNR using the proposed algorithm.
Visual Evidence
Fig. 2. MODWT decomposition showing (a) input signal, (b) first sub- band, (c) second sub-band, (d) third sub-band, (e) fourth sub-band, and (f) reconstructed signal from first three sub-bands. ECG R-Peaks shown in red dashed lines.
Clinical Snapshot
Evidence Rating
Relevance
high Priority