Re:Real-time tracking of human eyes using a webcam
by jauco on Wednesday August 27, 2008 @03:48AM
(#287853)
I know from using the tobii that eye-tracking has one big problem: Your eyes never stay still. Whenever you're focusing on something, you are making mini-movements (called micro saccades IIRC) around the focus point. When you move to the next point, you're eyes 'jump' there and do the mini-movement again. When you read a line of text, it feels as a fluid movement, but it actually is a series of jumps over a couple of words at a time.
Interpreting eye tracking is mostly defining what is the cut-off point between movement during focus and actual movement. It seems that with a web-cam, this choice is already made for you, due to it's low resulution :). It's been a year or two since I had to deal with it though, so my memory is a bit hazy. I can post links to some good articles on eye tracking if anyone is interested (just leave a reply).
Finally, I'd like to remark that many people put their mouse where their focus is. Mouse tracking is a quite effective (and cheap, and hardware-less) means of focus logging. As is simply recording clicks.
Re:Real-time tracking of human eyes using a webcam
by a-priori on Wednesday August 27, 2008 @10:09AM
(#288026)
You're dead on. The purpose for the micro-saccades is to keep the image on the retina changing. Photoreceptors on the retina do not respond to the intensity of light, but rather the rate of change of intensity. A combination of micro-saccades and blinking lets your visual system gather as much information as it can, and visual memory fills in the gaps.
The illusion of a complete image is really one of the most remarkable things about the visual system.
Anyone working in computer vision can tell me if I'm right or wrong? I've been working in the area in some small projects, but they were extremely related to this eye-tracking stuff.