Hi there, I will try to describe as simple as possible.
I have recorded video samples of hand movement. On every sample there is a black background and hand in the foreground. Every sample lasts for about 0.5-1 second.
I want to identify person by his/her hand movement, by analising it's optical flow by neural network.
I can compute optical flow for every frame, but I haven't found smart way of getting those information and passing it to ann_mlp. I was thinking about optical flow histograms, but I don't know if it will be successfull. Anyone has an idea?