Change the way activity for one image is integrated

[ecobost]

Until now, we used a hamming window of size 1 sec to integrate the activity. This uses a rectangular window over the half a second of presentation. Because we record at relatively slow fps (6-10), half a second of recording will only have 3-5 samples and using a weighted window (like hamming) will waste a good bit of info.

[fabiansinz]

The changes you propose would break all previous processing methods. Why did you decide against using a simple boxcar filter instead of the hamming filter? Wouldn't that have achieved the same thing? You trapezoid integration seems more complex than it needs to be.