If the whole setup can be built for $350 (which isn't unbelievable), then they aren't giving presentations at TED because they spent $350 on a webcam and projector. Writing software that can recognize faces is hard. Writing software that can recognize gestures is hard. He's written software that can distinguish between faces, gestures, and objects, and *then* recognize them. At least, that's the idea. I would love to see his project released as open source, but he has to resist the buy-out offers from major companies that he is surely getting. That is, of course, if the university doesn't claim ownership first.