Fusing Partial Camera Signals for Noncontact Pulse Rate Variability Measurement.

Area of Science:

• Physiological monitoring
• Computer vision
• Signal processing

Background:

• Remote photoplethysmography (PPG) enables noncontact physiological measurements.
• Head motion and occlusions cause data loss, hindering accurate pulse rate variability (PRV) calculations.
• Existing methods struggle with missing data from moving subjects.

Purpose of the Study:

• To develop and validate a robust method for noncontact PPG measurements from moving subjects.
• To fuse data from multiple cameras to mitigate the impact of missing observations.
• To improve the accuracy of pulse rate and PRV estimation in the presence of head motion.

Main Methods:

• An array of cameras captures partial color-channel signals.
• Fusion of signals from multiple cameras to reconstruct physiological data.
• Systematic testing on 25 subjects across various head motion tasks (25 hours total).
• Evaluation of pulse rate and PRV parameter estimation accuracy.

Main Results:

• Median absolute error in pulse rate measurement was 0.57 BPM.
• Error in pulse rate was within 0.4 BPM of contact PPG devices for most tasks.
• PRV estimation error was reduced by over 50% compared to methods not handling missing data.

Conclusions:

• The proposed multi-camera fusion approach significantly improves the accuracy of remote physiological measurements.
• This method enables reliable pulse rate and PRV estimation even with substantial head motion and data loss.
• It overcomes limitations of single-camera systems and enhances the feasibility of noncontact physiological monitoring.