error accumulation

[GoogleCodeExporter]

What steps will reproduce the problem?
1. run the 6dofhead
2. produce abrupt head motion
3. head angles will accumulate errors

What version of the product are you using? 0.7 
On what operating system? Win7

Any plans on implementing re-initialization to deal with error accumulation as 
you outlined in "next step" in wiki?
Any other ideas on dealing with this? I'm using it with Kinect now, and while 
it's beyond the scope of this project maybe the depth-map there can be used as 
an additional input modality? 
http://www.youtube.com/watch?v=MinkGUl2WnA

Original issue reported on code.google.com by alexleykin on 8 Mar 2011 at 4:58

[GoogleCodeExporter]

That's an interesting video.
Well, it seems error accumulation is the drawback of this approach. A simple 
idea would be leave Viola Jones face detector running and re-initialize the 
system whenever that happens. But I've been studying different approaches, a 
couple related to Active Appearance Models, and others related to 3d model 
tracking. I believe they will solve this problem, but in the long run.
Congratulations for your project. What are you using Kinect for in this demo?
Best regards,
Daniel

Original comment by danielba...@gmail.com on 8 Mar 2011 at 5:18

[GoogleCodeExporter]

Daniel,

I'm using Kinect to track joint angles of the body, but it's missing
head orientation, so I decided to use ehci for that. When you are
talking about running Viola Jones, do you mean assuming canonical
orientation, or am I missing something?

Alex Leykin

Original comment by alexleykin on 9 Mar 2011 at 2:57

[GoogleCodeExporter]

I see, that's great!!!

Well, when talking about Viola-Jones I'm assuming frontal face, canonical, for 
sure. 

Kind regards,
Daniel

Original comment by danielba...@gmail.com on 9 Mar 2011 at 3:14