Heart rate estimation from facial photoplethysmography during dynamic illuminance changes

Area of Science:

• Biomedical Engineering
• Signal Processing
• Computer Vision

Background:

• Camera-based remote photoplethysmography (rPPG) offers non-contact cardiovascular monitoring.
• Illuminance variations, especially from video content, create artifacts that limit rPPG accuracy.
• Existing rPPG methods struggle with dynamic lighting conditions common in real-world scenarios.

Purpose of the Study:

• To develop and validate a novel artifact reduction method for rPPG signals.
• To mitigate the impact of illuminance fluctuations on facial rPPG measurements.
• To enhance the reliability of non-contact heart rate monitoring in variable lighting.

Main Methods:

• A new artifact reduction technique was proposed for rPPG signals.
• The method utilizes multi-order curve fitting to correlate facial signal illuminance with video brightness.
• Artifacts are subtracted from the raw rPPG signal using this fitted curve.

Main Results:

• Significant improvement in signal-to-noise ratio (SNR) was observed, increasing from -11.74 dB to -4.19 dB (low brightness) and -15.27 dB to 7.99 dB (high brightness).
• Root-mean-square error (RMSE) for estimated heart rate decreased substantially, from 11.00 bpm to 1.82 bpm and 9.88 bpm to 4.65 bpm.
• The proposed method effectively reduces artifacts caused by dynamic illuminance changes during video viewing.

Conclusions:

• The developed method successfully reduces illuminance-induced artifacts in rPPG.
• This technique enhances the accuracy and robustness of non-contact heart rate monitoring.
• The findings suggest improved practical applicability of rPPG in diverse lighting environments.