The raw image capture tut is the one I think. Code is included... All in C++ of course...
http://software.intel.com/en-us/articles/intel-perceptual-computing-sdk-tutorials
The depth image looks pretty good, with a little smoothing it looks useful. 320x240. multiple samples froma moving camera should increas the effective accuracy...
Do you not think so ?