electrooculographic (EOG) eye tracking

That an interesting one, although it is a bit awkward looking. (as these kind of things usually are... XD)

The Emotiv Epoc headset can supposedly track facial expressions from EEG readings.
I don't know how well it does at tracking eye movement, but I assume it tracks facial expressions using similar principles.
(Altough it can also track emotional states too, which is interesting to think about...)

Another one that might be of interest is the Tobii system they showcased at CES this year:

http://www.tobii.com/en/gaze-interaction/global/solutions/devices-components/tobii-is20-tobii-is30/

http://www.tobii.com/en/gaze-interaction/global/eye-tracking/

These have been on my TODO list for reading for a bit, probably a good time to dig in and add those others.

EDIT:
Hmm, looking at that closer, I'm not sure how you'd set that up with the eye cups in place.

"Felix12g" wrote:
Another one that might be of interest is the Tobii system they showcased at CES this year:

http://www.tobii.com/en/gaze-interaction/global/solutions/devices-components/tobii-is20-tobii-is30/

http://www.tobii.com/en/gaze-interaction/global/eye-tracking/

These have been on my TODO list for reading for a bit, probably a good time to dig in and add those others.

EDIT:
Hmm, looking at that closer, I'm not sure how you'd set that up with the eye cups in place.

EXACTLY! Your eyes need to be visible to external eye-tracker cameras. That is why EOG is so interesting for use with a Rift DK -- the eyecups and electrodes can coexists without difficulties.

I wonder if you could get away with positioning a small camera inside the screen chamber ( or whatever you call it ) on the Rift such that it could capture the warped orientation of the eyes, but be visually ignored in the same way that your brain hides your nose from your vision. You might not even have to un-warp the images to do the necessary processing to calculate eye orientation.

"nsb" wrote:
I wonder if you could get away with positioning a small camera inside the screen chamber ( or whatever you call it ) on the Rift such that it could capture the warped orientation of the eyes, but be visually ignored in the same way that your brain hides your nose from your vision. You might not even have to un-warp the images to do the necessary processing to calculate eye orientation.

I put some thought into that arrangement, using beam splitters between the screen and the lenses, but viewing in the reverse direction through the lenses causes even more distortion and would probably not work well (if at all) with existing eye-tracking software. You may possibly be able to use complex optics to undistort the image of the eyes, of some kind of software unwarping.

A problem with a beam-splitter is that a small one will be visible (reflecting part of the light away from the eye), and if large enough to cover the full FoV so no edges can be seen, it would be difficult to fit between the lens and screen (unless the cameras were mounted right next to the eyes).

EDIT: Perhaps a beam-splitter INSIDE the eyecup (with a camera just outside the eyecup) may be worth a try.

What about a bracket that lays down over the eye cups? Take the IR illuminators on adjustable rails that lay down over the top of the cups, you can keep the wiring and orientation stable with that. I'm guessing the cups aren't so close that your eyelashes are brushing the lenses when you blink?

My other question was whether the eye tracking cameras need to see both eyes, each, or if a single eye per camera is fine.

One thing that caught my eye in the EOG paper right off the bat (And no, I've not finished it all) was their mention that long term use of IR illuminators on the eye hasn't been properly tested, so that's a concern.

"Felix12g" wrote:
What about a bracket that lays down over the eye cups? Take the IR illuminators on adjustable rails that lay down over the top of the cups, you can keep the wiring and orientation stable with that. I'm guessing the cups aren't so close that your eyelashes are brushing the lenses when you blink?

My other question was whether the eye tracking cameras need to see both eyes, each, or if a single eye per camera is fine.

One thing that caught my eye in the EOG paper right off the bat (And no, I've not finished it all) was their mention that long term use of IR illuminators on the eye hasn't been properly tested, so that's a concern.

As I have mentioned in many of my posts, my eyelashes DO brush the lenses. I place the lenses that close to my eyes because it significantly increases my FoV. There is no room for anything substantial between my eyes and lenses.

I have also posted previously that you can use a single camera for stereoscopic content if you assume that both eyes are converged on the corresponding pixel, which can be determined by software.

Great resources, thanks.

"geekmaster" wrote:
Perhaps such electrodes could be attached to or embedded in the Rift facemask foam....

My thoughts exactly.

I came across this HackADay article, Tracking eye movement by measuring electrons in the eye, a few years back when thinking about sleep masks to read REM state.

When I learned of the Rift, I started wondering about the possibility of combining it with something like your brainwaves, and Binaural beats as a Lucid Dreaming training aid.

I also intend to experiment with vision correction (myopia in my case) in software. With enough information about the eyes: IPD, eyeglass prescription (ideally done objectively, rather than with a phorepter), and eye tracking, it may be possible to present a distorted image that winds up corrected by the viewer's eye-condition.

Originally, I was thinking that a vision-aware window manager could automatically correct the view based on the user's distance from the screen via camera tracking (OpenCV), which would also track eye position (obviously).

Edit: hah, I just noticed that your arduino image is taken from a Lucid Dreaming arduino project.

This video shows just how effective EOG can be:



Just imaging that level of control in the Rift DK HMD, and how you could use that to optimize rendering and other uses...