I did this because its very fun and innovative and I've never tried something like this before. It was very fun but confusing and I haveto admit many times there were difficulties and challenges. 

It works because the AI sees the gestrues and connects them with the command and what it learns is that a certain gesture meant a certain thing. So it collected the many images of lets say my fist and understood that when I put my fist up it meant to go to the previous track.