Exploiting spatial redundancy of image sensor for motion robust rPPG.

Area of Science:

• Biomedical Engineering
• Computer Vision
• Physiological Monitoring

Background:

• Remote photoplethysmography (rPPG) measures cardiac activity via skin color changes detected by RGB cameras.
• Existing rPPG methods struggle with motion artifacts, leading to inaccurate pulse rate measurements.

Purpose of the Study:

• To develop a novel rPPG framework that improves robustness against subject motion.
• To leverage spatial redundancy in camera imaging for enhanced pulse signal extraction.

Main Methods:

• Proposed a pixel-based rPPG framework utilizing motion-compensated pixel-to-pixel pulse extraction.
• Implemented spatial pruning and temporal filtering to isolate pulse signals from noise.
• Evaluated the method on 36 benchmark videos with diverse subjects and motion types.

Main Results:

• Significantly improved signal-to-noise ratio (SNR) from 3.34 dB to 6.76 dB compared to state-of-the-art.
• Increased agreement with reference pulse rate from 55% to 80% (±1.96σ).
• Demonstrated significant improvement in motion robustness via ANOVA testing.

Conclusions:

• The developed rPPG method achieves high accuracy comparable to contact-based sensors in realistic scenarios.
• The framework's computational efficiency enables real-time processing on standard hardware.
• This approach offers a promising non-contact solution for physiological monitoring.