The evaluation of seismocardiogram signal pre-processing using hybridized variational mode decomposition method

Area of Science:

• Biomedical Engineering
• Signal Processing
• Wearable Technology

Background:

• Single-channel accelerometry data often contains noise, complicating the extraction of vital physiological signals.
• Seismocardiogram (SCG) and respiration signals are crucial for cardiovascular and respiratory monitoring.
• Existing signal processing methods struggle with noise interference and accurate signal separation.

Purpose of the Study:

• To evaluate a novel hybrid framework combining Variational Mode Decomposition (VMD) and Detrended Fluctuation Analysis (DFA).
• To extract seismocardiogram (SCG) and respiration signals from simulated noisy single-channel accelerometry data.
• To assess the framework's performance in noise reduction and signal fidelity.

Main Methods:

• A two-layer VMD approach was employed for sequential extraction of respiration and SCG signals.
• DFA was used to determine the optimal number of VMD modes via Hurst exponent thresholding.
• Signal reconstruction was performed using selected VMD modes.

Main Results:

• The hybrid VMD-DFA framework successfully extracted SCG and respiration signals with low mean absolute errors (0.516 and 0.849).
• Signal-to-noise ratio (SNR) was improved by 2 dB for respiration and 4 dB for SCG signals.
• The method demonstrated superior performance compared to other empirical mode decomposition techniques with reduced computational time.

Conclusions:

• The hybridized VMD-DFA framework offers an effective solution for denoising and extracting SCG and respiration signals from combined, noisy accelerometry data.
• Despite minor drawbacks like manual parameter tuning and magnitude shifting, the method shows outstanding performance.
• This approach holds promise for non-invasive physiological monitoring using wearable sensors.