Project DetailsThis work introduces a system that estimates a patient’s heart rate by observing changes in light reflection from their skin under near infra-red (NIR) light. The heart rate (HR) estimation is based on a photoplethysmography (PPG) signal that is extracted using a NIR camera and analyzed in the frequency domain. A pre-trained facial recognition model was employed to detect the subject’s facial skin region in the video. In each of the video’s frames we identified different regions of interest (ROI's) based on known anatomical symmetries. Each ROI was averaged and given weight, the weighted average values were summed to generate a 1D time-series. The signal’s spectrogram was observed, and the frequency with the highest median value in the physiological HR range was estimated to be the HR.
After optimizing the system’s parameters (window length, window overlap, regions of interest’s weights), the resulting RMSE was 4.31𝑏𝑝𝑚 and the median error was 4.09𝑏𝑝𝑚. Our system was tested using subjects from various ethnicities (Israeli, Indian, Chinese), demonstrating robustness and reliability. While the method has a limited frequency resolution due to the sampling frequency, the results could be further improved using cameras with a higher sampling rate.
