Dutch company Softkinetic offers what they call natural interfaces, i.e. interfaces where you don’t have to put on any sensors to interact:

Softkinetic operates with a single depth sensing camera, requires no marker (no gamepad, no wiimote, no special gloves or clothing, no headset – nothing), and works under all lighting conditions and scene settings (at home, in a fitness center, an amusement park, a classroom, a game cafe, an industrial simulation room – anywhere.)

Using a depth sensing camera obviously greatly simplifies the computer vision, since it allows for much simpler foreground/background subtraction. It would be great to hear from people that have experience with such 3D cameras, and hear which models work well.