Yet another experimentation with a Kinect unit, Skeleton recognition and video mapping.
The system is working so far, but there’s an annoying one frame delay that forces to move slowly. Video in the making.



Like this:
Like Loading...
Related
This entry was posted on March 9, 2011 at 14:06 and is filed under experimental with tags dancing, kinect, openGL, video mapping. You can follow any responses to this entry through the RSS 2.0 feed.
You can leave a response, or trackback from your own site.
March 9, 2011 at 19:15
hi franz could it be the usb cable length?
March 9, 2011 at 21:00
Tu sais que j’adore danser… .-))
March 9, 2011 at 23:16
Using the standard cable….
Maybe it is my setup but I doubt I can do anything about it.
On the mac you don’t notice the 1 frame delay, but when projected over the real action, you do .. :.(
March 9, 2011 at 23:29
then maybe a too long video cable, back in analog days I would do a video delay with a 50mt cable rolled up next to the vcr going to the mixing desk and to the projector :) cracking 1 frame
March 10, 2011 at 01:15
Looks awesome man! congrats.
Do you have the delay on just the projection? or also on screen? Projectors can have upto 100ms delay (at 30fps thats 3 frames!). The best consumer projectors I know of have 16ms (1 frame at 60fps). You can get specialist projectors with 2/3ms but they’re quite expensive, and not very powerful. There are projector latency measuring tools out there, but you can knock one up quite easily in QC. just project a timecode, have it run on your monitor and screen at the same time, then take a photo of both monitor and projected surface, see what your projection latency is.
March 10, 2011 at 10:45
thx memo, will look at this technique. 1 frame delay is really annoying for fast movements…. Hope to get this sorted
March 10, 2011 at 12:35
ok, just tested my projector latency, and effectively, there’s a 16ms delay on the projector. 1 frame … boooo
March 10, 2011 at 12:50
yea very annoying, but 16ms is the lowest latency I’ve seen in a projector at consumer level! It’s all that digital image processing, keystoning, color correction and other crap, I don’t think you can get passed it.
March 10, 2011 at 16:44
well you could allways adapt a “silent movies” dance style. :)
March 30, 2011 at 04:31
sounds crazy, but perhaps you can do motion prediction to predict where skeleton nodes will be 3 frames from now. this will work well for consistent movements, and as long as you keep your predictions 1st order (predict velocity, not acceleration), then you shouldn’t be any worse than you are now with rapid changes in movements.