Just to play devils advocate, you don't want just data, you want some interpretation of it as well. Kinect (at least the original, not sure about the new one)

Just to play devils advocate, you don't want just data, you want some interpretation of it as well. Kinect (at least the original, not sure about the new one) does not do on-board joint identification. You just get depth data. You want data and the various extremely advanced interpretations of it, something that is not trivial to build on your own. So there needs to exist at least one more processing layer before you can consume it for most applications. http://en.wikipedia.org/wiki/OpenNI http://en.wikipedia.org/wiki/OpenNI is a good step, but until those other layers exist reliably in a stable cross-platform way, "just the data" is nearly useless.