How Natal Works
The test system was an ordinary Xbox 360, connected to small PC and camera that simulates the final Natal rig. There are two cameras—one RGB, for face recognition and display video, and one infrared, for tracking movement and depth. Why infrared? The eye doesn't see infrared light. And when you combine an infrared camera with an infrared emitter (also part of Natal), a room is flooded with a spectrum of invisible light that works in the dark.
Natal also has its own internal processing system handling an unspecified amount of the heavy lifting behind Natal's cleaver image and speech recognition. It breaks the human body into 48 points tracked in real time, and it can sense your whole body in Z space, or depth. In fact, on a heat map that measured depth, my hands appeared hotter than my shoulders—because they were closer.
Natal is so smart, in fact, that, if your room is narrowed by a pair of couches, it can signal to a game to narrow the level. It can see about 15' x 20' of a room, according to project leader Kudo Tsunoda's informal estimation.