@TanMath I am getting a bit confused now. 1) Take a few measurements with no input (i.e. just a picture without any periodicity). This will give us the background "noise", or in other words, the activity that the system generates with zero input. 2) Take a few measurements WITH a periodic input WITHIN the frequency range. (i.e. ~1, 1.5Hz) This will tell us what sort of input is required to generate a valid output and also will tell us how quickly the system responds to changes.

@TanMath The camera doesn't look like it has auto-exposure or if it has it is well hidden or you cannot control it. Cheap web cams (and I mean, cheap and not bad) go for under £10 in Amazon (or other convenient retailer). See for example this one: amazon.co.uk/Logitech-960-000355-Webcam-C250/dp/B002CNERDE/… or this one

amazon.co.uk/Logitech-961237-0914-QuickCam-Messenger-Webcam/dp/…

@TanMath I have a couple of the latter model, they are great generic web cams.

@TanMath I am not sure that the camera does auto exposure but it definitely has flicker compensation as you can see. It is now set to high quality, which I would assume does a bit of processing on the image. So, let's try "Speed Priority" and hope that in that setting it passes frames with minimal processing.

@A_A periodic is the input from the webcam? cuz I already did that and gave you the results

@TanMath The image settings controls seem disabled (?). See if they become enabled once the video priority is set to speed.

@TanMath Hello, I don't understand what do you mean with "From the webcam"?

@A_A background I did one in various backgrounds and got oscillatory signals... but you want one with the picture...

@TanMath I am getting even more confused now.

@A_A like for example i also am doing with only the normalization+ICA+bandpass+FFT with synthetic signals on my computer...

@TanMath To avoid any confusion can we please agree to a plan?

@A_A sure I am so sorry for the confusion...

@TanMath No worries. Here is the plan:

Or at least, what I would suggest

1) Take a few measurements with no input (i.e. just a picture without any periodicity). This will give us the background "noise", or in other words, the activity that the system generates with zero input

Just cover the camera with your hand

No input at all

@A_A but there's periodicity in my hand (blood flow)

Cover it with a few pieces of black paper so that no light shines through (?)

hmmm I can try to do that ok...

After that. Do the experiment with the picture. Again, NO input

Just a plain picture.

2) Take a few measurements WITH a periodic input WITHIN the frequency range. (i.e. ~1, 1.5Hz)

For these two experiments, please post:

ok I have gifs prepared in that range..

1) The spatial averages WITHOUT any processing. No normalisation no nothing

2) The normalised spatial averages. For this step please mention EXACTLY how you apply the normalisation because this step might be injecting a signal

That is where we are going to start from

I am in the same timezone, I was traveling in the weekend and I will be traveling again tomorrow and the day after, so there might be delays. It is now 09:21 local time

@A_A for the normalization I use the scikit-learn "standardization" functionality...

        G_std = StandardScaler().fit_transform(np.array(G[-(N-1):]).reshape(-1, 1))
        G_std = G_std.reshape(1, -1)[0]

like that for each color channel

anyway I will post those experiments tomorrow...

@TanMath, please turn that off

@A_A I assume you do not have time in the night though? so like tomorrow same time do you think you would have a chance to look at the experiments?

@A_A ok turn it off for part 1 but for part 2 you want it on?

I can check things relatively easily, but I am not sure if I will be able to respond, whether because I am busy with other things or because the network is playing up

No, turn this scaling off in general. For the above, please do not produce ANY plots for the ICA part. We will bring that gradually in later

In brief, if you don't get a decent signal in the input, there is no point in using the ICA bit

@A_A hmmm? I am confused then what to do for part 2...

@A_A ok makes sense will do... (i.e. remove the ICA plots )

@TanMath, the point is to examine if your webcam setup responds to your signal

@A_A yeah but do you want this?:

because then i need to turn on the standardization function

@TanMath ah, i see. No, this has been answered. For both steps, please have the normalisation turned off

ok to confirm:

1. black image --> spatial avg. no normalization

2. image of my face --> spatial avg no normalization

3. pulsating gif --> spatial avg. no normalization

correct?

@A_A ^

For #2, PICTURE of a face

@TanMath ^

@A_A ok like a printed out picture of a face (or on another computer screen fine?) in front of the webcam recording for like a minute...

actually for all of them, correct?

@TanMath Yes

ok then...

@TanMath Your pipeline seems to be Webcam->Face recognition->Spatial average->Bandpass filtering->ICA->FFT

@A_A ICA and bandpass switched...

@TanMath Even worse :D

@A_A why?

@TanMath I think that the bandpass is ringing, but we will look at this more carefully when we get there

@A_A oh right you mentioned that before ok...

@TanMath that image is very useful, thank you very much

no problem...

@A_A how much time do you have right now?

cuz i can try to get the first experiment to you in like 10-20 min

@TanMath I might be able to pass some comments to you but I think that it would be best if we did things in "Plan->Action->Review->Plan->Action->Review"

@TanMath otherwise it gets too confusing.

@TanMath BTW, what language are you using for this?

@A_A so like all of the results at once?

@A_A python

@TanMath OK, that's good to know

@A_A which reminds me, the code was posted in the question, do you think it is worth for you to run it? probably after i finish these tasks you said first though...

I can have a quick look but that requires more time than I can devote to this for the moment.

@A_A this is fine with me... I know you are probably extremely busy and you are already helping me so much but the 24 hour delay kind of makes it hard for me... I am under a time constraint to finish this up... it needs to be uploaded to raspberry pi and working in like two weeks, so code should be ASAP ... I understand that you are very busy, but I would greatly appreciate it if you have other chances to look at it...

@A_A so i guess not, huh? is there anything else I could do to, i dunno, like streamline the process or something? I guess anything on my side to help save time?

@TanMath I am not keeping my eyes on this screen all of the time. I am not ignoring the messages. I can help you with the technical part as much as I can given the current circumstances but I do not see how I can help further with the rest of the issues, such as deadlines and time pressure.

@TanMath

@A_A ok i understand... is there anything I should do on my side though?

@TanMath I would recommend that you don't stress yourself too much because it affects your attention and ability to deal with the situation. Please don't ignore to take some rest

@TanMath Port the code to the Raspberry pi and work there so that you don't feel that you are wasting any time (?)

@A_A it was planned that the code would need to be optimized for the raspberry pi first but yeah maybe I will do that given the time circumstances...

@A_A well thank you for the help so far! i will get those results to you tomorrow (like probably 12-15 hours from now)... talk to you later...

@TanMath All the best