well it may not be new on the techonology area, but they are two expensive and the chipest normally are too slow or do offline calculations, so this year i started a project with other physics people in a way of developing an eyetracker that is fastest, chipiest and does not require too much hardware... At this stage we have create several algorithms... I will post some C code when i get it to work online at this stage we are only in the creation of the algorythms as the test in some images.... post your thoughts

I've seen several eye tracking programs that work at an acceptable speed without any trouble....

Just a question; I'm not too keen on digital imaging and the processing of it, but wouldn't a pupil movement detection algorithm be as simple as detecting the border of the eye using average pixel distribution, detecting the pupil using more average pixel distributions and/or user-supplied eye color data and adjusting for rotation by anamorphically distorting the image?

Correct, but that's way, way more complicated to do in a CPU-conservative manner than you'd think.

Would glyph-tracking work for this, if the glyph being tracked was your eye? Or is that the memory intensive way...

Glyph tracking won't work; the camera needs to capture a unique glyph it is programmed to recognize. Perhaps in the future as CPU's get faster and the computer/program will be able to recognize partial glyphs or fractals such as those in the eye - but camera quality would have to be above par to capture such detail.

The eye is as unique as the fingerprint - on the retina at least - and the iris is no subsitute for a glyph since it will slightly vary from person to person.

The best way for glyph tracking is to wear a contact lens(es) with a glyph printed on that the camera can recognize with ease.

Check out: http://youtube.com/watch?v=pnDZsXK1X9Y

sorry for the long term off but i have been working on the subject on this months and to answer kll yes there are some but they are not cheap, and many of them require lots of hardware what we attempt to accomplish is something that is fast cheap and requires only a camera and a computer...

About the algorithm i can say that we track two things the pupil center and the glint center, based on ellipse fits!

when i get it finished i will post here all code, as also the conference paper!!!

Very nice, what conference are you publishing in?

Wow! Good job, so you have it working?

Can't wait to read the conference paper ;D
Is posting the code, publicly, a good idea?

Yes i have it working!
At this moment no! After the conference we decide to publish to all view in the hope thatit gets futher developed!