Monday, 8 November 2010

Kinect Gets Hacked: More Hardware Info Surfaces

Getting console peripherals fully up and running on the PC has been pretty much commonplace over the last few years. In that time we have seen certain individuals getting both Wii Remote and Nunchuck compatibility in Half-Life 2, whilst others have plundered the PSEye, creating new drivers to exploit the device’s full functionality. Now someone has seen fit to do the same thing with Microsoft’s Kinect, albeit in just a few days.

Yesterday it was reported over at gizmodo that the NUIGroup had successfully managed to hack Microsoft’s Kinect, and have, in just a short space of time, coded custom drivers for the hardware enabling it to work on PC’s. Whilst skeletal tracking currently isn’t available (this is done through MSs own software libraries) the person behind the hack has demonstrated both motor and accelerometer control, in addition to having full access to audio and visual feeds from the camera. Videos can be seen here and here.

Interestingly, control over both the both the motors is done by the host platform – either the PC or the Xbox 360, with no control over user tracking. Which begs the question of whether or not it is the Kinect itself is actually doing the tracking we originally though the 360 to be taking care of. At the moment this isn’t clear, although what is somewhat revealing, thanks to an iFixit teardown, is that the final retail version of Kinect does still have an onboard PrimeSense processor.

Could the device still be performing some of the tracking work? If so, then that would explain its presence, although not why the 360 has to perform the actual calculations for skeletal tracking, and not the simple user tracking as talked about here. Then again, some have alluded to an additional MS-based processing chip having been removed, which was said to have handled that particular task.

The inclusion the processor aside, we do know that while Microsoft’s own documentation lists the depth feed as 320x240, the actual hardware contained inside still has the capacity to record it in full 640x480 resolution. But why the apparent restriction? Again, we simply don’t know at this point, although more accurate tracking could be possible if developers have access to the full res feed.

Perhaps MS’s own API is what restricts this? We know for a fact that developers have openly stated in the past that it is the API that restricts exactly how Kinect can be used, and that future updates should expand functionality further. So this certainly wouldn’t be beyond the realm of possibility.

From what has been uncovered so far, it is pretty clear that whatever cutbacks have been made in getting Kinect out as a viable retail product, much of the original PrimeSense tech remains solidly intact, and perhaps even underused. The revelation of full 640x480 depth buffers for example, is reason enough to believe there is more developers could be doing with the hardware. And the discovery of a 3-axis accelerometer in the rotating camera itself, rather outworldly, allows the same tech to be adapted for use in robotics, with the accelerometer being used in orientation and depth-sensing for head movement.

One thing is for certain though, that NUIGroup’s reverse engineering of the Kinect, making it work on a PC, is just the beginning. We’re sure to find out far even more about the device in subsequent weeks after more tests have been done. Until then, this is just a mere taster of things to come.

No comments:

Post a Comment