I'm finally making my way back to RoF after a few busy months and decided a head tracker would be nice. While I'll probably end up buying a TrackIR, since it seems to be the gold standard, I wanted to give the DiY alternatives a try first.
FaceTrackNoIr (http://facetracknoir...ome/default.htm) didn't cut it, then I ran across EDTracker (http://www.edtracker.org.uk/index.php) which looks very promising. It was initially done for Elite Dangerous (space-sim game) but has since been used elsewhere. It uses a 6 (or 9) DoF accelerometer chip along with the same Arduino pro-micro chip used by the MMJOY2 project HippyDruid was talking about. Cost to build is about $20 US for the Arduino and the mpu-9150 (or 9250). Even the pre-assembled 'Pro' version sells for under $50 (I think).
So anyway, I did build it -- pretty easy, however it's not working with RoF. In the GUI made by the developer, it tracks my head movements pretty nicely (straps to the top of headphones with a micro USB cable attached).
It also looks good in the Windows USB Game Controller test window. But in RoF, while the 2 axes are detected and assigned for camera control, in the reponse curve section, it shows NaN (not a number) indicating some mathematical error (like a divide by zero). Using the windows joystick calibrate function, I did notice that the raw output values coming from the EDTracker HID joystick are going from -+/32,768 (e.g. a 16-bit signed integer). I'm guessing this is waaay higher resolution than an average joystick and may be screwing things up.
Anyone ever try this before? Or even hear of it? I expected there to be a lively debate in the forum archives, but not a mention.
FYI, there *is* a way to get EDTracker to look like TrackIR with a program called 'opentrack', so that's probably my next thing to try.