I'm working on a project where I have to detect hands clapping - ideally where the hands meet; the approaches I've tried thus far are: - getting the centre of gravity from the last frames mask (using the absolute difference from an accumulated background image)
centre of gravity on a mask based on skin thresholding
optical flow; I group all the features based on distance and work out the average distance between the 2 largest groups.
(2 example images shown below).
The user can clap anywhere (ie not in a controlled environment/situation).
Any suggestions, hints, tips, solutions would be greatly appreciated,
Josh
