Inaccurate at high heart rate

[connormeaton]

Thanks for providing your code!

It is very accurate at low BPM levels. However, when I raise my heart rate, the accuracy decreases.

For example, if I do no exercise, my bpm is accurately predicted at around 60 BPM (with good lighting and no movement). If I then exercise and get my BPM up to 120 (verified by a wearable oximeter) and then predict my bpm with your code again (with the same good lighting and no movement), the code predicts in the 60-80 BPM range. This is wrong by 40-60 BPMs. This seems to be a pattern across rPPG algorithms. Do you have any idea why this is the case? Thanks!

[MartinChristiaan]

Hi, thanks for your interest.
I am afraid I cannot explain the reduced accuracy, are you sure your framerate is sufficient? I never did a test with an oximeter myself but I was able to measure an elevated heartrate >100BPM after going for a run.

[connormeaton]

Thanks for the response! I don't think framerate should be a problem. I'm running between 25-30 fps.

What framerate did you use to measure >100BPM?

Also, was the >100BPM measurement consistent throughout the duration of the video/stream, or did the BPM value jump around a lot? For example, if you were to take the mean BPM of that video/stream, would it be >100BPM?

Thanks!