Reliability of Color-Based Pulse Detection Heart Rate Variability Analysis Software
When a human heart pumps blood, there are slight color changes in the forehead. These changes are unobservable by the naked eye but are detectable through cameras. Color-based pulse detection records the color change of a video subject’s forehead and outputs the RR intervals (time intervals between subsequent heartbeats); however, video quality can interfere with this process. One aspect that affects video quality is file compression. To test the reliability of color-based pulse detection and its dependence on the file format chosen, I examined 5 file compressions (.avi, .mov, .mpg, .wmv, and .mp4) on 5 videos of a light-complexioned subject and a dark-complexioned subject and compared the mean heart rate computed using color-based pulse detection with RR intervals generated from the Pan-Tompkins algorithm of electrocardiogram (ECG) signals on the subjects. For the light-complexioned subject, the .mpg compression yielded the highest correlation with the ECG data (r squared = 0.953). For the dark-complexioned subject, it was found that .avi yielded the highest correlation (r squared = 0.812), a significantly lower coefficient than that of the light-complexioned subject. This indicates that while compression seems to have a minimal effect when starting with a high signal-to-noise ratio (light-skinned subject), it does deteriorate algorithm performance at a lower signal-to-noise ratio (dark-skinned subject). It is important to know the consistency of this method at different levels of video quality and compression. However, this research points to the fact that compression may challenge video-based heart rate detection in darker complexioned people.
Copyright (c) 2019 Karen Song, Vasiliki Ikonomidou
This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.