Robust Methods for Automated Selection of Cardiac Signals After Blind Source Separation

Area of Science:

• Biomedical Engineering
• Signal Processing
• Wearable Health Technology

Background:

• Minimum-contact vital signs monitoring, including textile and capacitive electrocardiogram (ECG), offers advanced health monitoring capabilities.
• These novel techniques are susceptible to motion artifacts and signal instability, posing challenges for accurate data extraction.
• Spatio-temporal blind source separation (BSS) can process multichannel signals but faces permutation indeterminacy, requiring cardiac signal identification.

Purpose of the Study:

• To evaluate different concepts for resolving permutation indeterminacy in spatio-temporal BSS for vital signs monitoring.
• To compare novel automated component selection routines based on heartbeat detection against standard BSS component selection methods.

Main Methods:

• Applied spatio-temporal BSS to textile and capacitive ECG datasets from subjects performing motion protocols.
• Utilized the MIT-BIH Arrhythmia Database for evaluating performance on arrhythmia data.
• Compared heartbeat-detection-based selection routines with standard methods (higher-order moments, frequency-domain features) using heartbeat detection accuracy (ACC).

Main Results:

• Heartbeat-detection-based selection routines significantly outperformed standard selectors (Skewness, Kurtosis, frequency-domain features), particularly with motion artifacts.
• For arrhythmia data, beat analysis using sparse coding showed superior performance compared to simple periodicity tests.
• The developed methods reliably selected cardiac signals from spatio-temporal BSS outputs in the presence of severe artifacts and arrhythmias.

Conclusions:

• Automated component selection routines based on heartbeat detection effectively resolve permutation indeterminacy in spatio-temporal BSS.
• These robust selectors enable reliable cardiac signal extraction from artifact-sensitive minimum-contact vital signs monitoring.
• Facilitates the broader adoption of advanced BSS techniques in wearable health monitoring systems.