Improved ultra-short-term pulse rate variability using hair detection and majority vote in subintervals

Remote photoplethysmography (rPPG) is a non-contact method for extracting pulse signal from a region of interest (ROI) in a human facial video. This technique enables researchers to remotely measure both heart rate and pulse rate variability (PRV). However, when the forehead is used as the ROI for rPPG signal extraction, hair can obscure parts of the skin, and changes in ambient lighting may introduce spurious frequency spikes, which degrade the rPPG signal and PRV accuracy. This paper proposed a method to improve ultra-short-term PRV derived from the rPPG signal using the forehead ROI. The approach incorporated a hair detection algorithm to extract the rPPG signal from skin areas, excluding regions covered by hair. In addition, a majority voting mechanism was applied to subintervals to determine the optimal passband frequency for a bandpass filter, effectively eliminating spurious frequencies. The ultra-short-term PRV was then computed from the refined rPPG signal. Results show that the mean absolute error of the ultra-short-term PRV was improved for most subjects compared to the mean absolute error obtained via the conventional method.