Playing with kinect
Lately i have been playing with kinect aiming having a working free software based head pose detection working for my phd thesis. The thesis is about spatial audio and in those set ups is extremely useful knowing the head pose. So, yes, I am an audio guy playing with computer vision so don’t hesitate correct me if I say something wrong or suggest me better approach than i took.
There are two competing projects supporting kinect.
The second comes from the the original company that developed the Kinnect. It has been partially opensourced but still the important bits such as pose detection are closed. Available for free as binary blob, which is a real headache in Linux because of binary compatibility. Furthermore, the API for me is too M$. I really didn’t like it. So my plan is to discard it as
The former, freenect or openkinect, is more open. It started with a competition to get basic drivers for the Kinect in Linux. It was an astonishing success, the result was the freenect library. But having that is just having a way to get the image from the IR and RGB cameras.
A lot of work is needed beyond that just to get where openni got. And then i have to get head pose detection. So this is the first of a series of post explaining my progresses implementing head pose detection with kinect which is part
Update: I had to stop the project but you can get the progress i made at github Take a look to this video!