@TanMath I am very glad to hear that. Can you please post similar plots of the PPG related signal?

@TanMath thanks, yes it was getting rather late and it's a full day today including traveling back

@A_A you wanted the ICA plots and the RGB plots?

@A_A here are the RGB, ICA, and FFT plots:

@A_A the FFT seems to be quite noisy, with a relatively strong peak still around 50 bpm.. in fact, sometimes, my program still prints around 50 bpm, but mostly prints around 70-74 bpm, which is close to my heart rate...

@A_A However, after doing a run, I ran the program, and it could not detect my high heart rate and was giving around 50 bpm as usual...

@A_A huh... I am also traveling today/tomorrow, what a coincidence... :)

@A_A I just did another run of the program and got this FFT:

@A_A I do think it might be due to the lighting as right now it is night... I have many lights on in my room but still, my earlier test I just posted was when it was daytime, with sunlight from the windows also there...

@A_A it seems from repeated experiments, that good lighitng and still face can improve SNR (as expected i guess), but the 50 bpm signal still occasionally shows up... I still want to figure where that is coming from and try to remove it and how to improve the SNR because other papers seem to have really good SNR, and cleaner FFT spectra...

@TanMath Can you please run the following experiment when you can:

@TanMath No ICA, no filtering

@TanMath of course, between 2 and 3 you extract spatial average from the face area

@TanMath Can you try changing the light flicker compensation from 60 to 50hz? if the peak shifts it is injected by the compensation algorithm

this is the regular one, not with (R+G+B)/3... the 50 bpm is still there...

@A_A I am at an airport, so the lighting is bad, and overall lots of noise ... but the signal is there at 1.2 Hz... but i guess the flicker might not be the problem...

@A_A ok so for the intensity experiment, I kept everything the same as the previous experiments (i.e. spatial avg. of each color channel, no normalization or bandpass), except after obtaining the RGB temporal signals, I averaged them as you specified, and did the FFT (so no ICA like you said)... I got weird results... for one, the heart rate signal did not show up at all:

So very weird...

@A_A given that I am doing this in an airport, with my laptop on my lap and not on a table, there were some sudden changes due to slight movement, travelers bumping into me, etc... so I guess take these results with a grain of salt? but they aren't that bad, so hopefully we can conclude something from these plots...