[Sigia-l] eye tracking?

[Steven Pautz]

> As part of this expansion, i want to persuade the company to invest in
> some eye-tracking equipment: [....]

> I've done a bit of googling, but i thought the list might have some
> personal recommendations for hardware to choose (rather than relying on
> manufacturers' sites). Also if you've got any eye-tracking war stories..
> all advice is appreciated.

I've heard good things about the Tobii -- http://www.tobii.com/ . The
eyetracker seems fairly accurate, and the built-in-monitor setup is
less distracting than some of the pan/tilt units. I haven't used the
bundled software myself, but it seems very powerful and flexible, once
you get used to it. It doesn't require a chinrest (horray!), but the
user has to keep their head fairly still nontheless. I had to remove
my glasses when using it, but I haven't used (or heard of) any
eyetracker which worked well with glasses. =)

If money happens to be an issue, the OpenEyes project at Iowa State
might be useful -- http://hcvl.hci.iastate.edu/openEyes/ . It's an
open-source, open-hardware system with plans for both a mobile
head-mounted eyetracker and a pan-tilt eyetracker. Both are intended
to be low-cost and easily-constructible. I don't know many details
about their accuracy or use, but lab head Derrick Parkhurst
(http://hcvl.hci.iastate.edu/cgi-bin/parkhurst.cgi) can probably
answer any questions, if you're curious.

As a reference, I'd recommend "Eye Tracking Methodology: Theory and
Practice" (ISBN 1852336668), by Andrew Duchowski
(http://andrewd.ces.clemson.edu/). It contains a lot of information
about eyetracking devices and the technology and science that goes
into them, as well as relevant information about experimental
methodologies and the human visual system.


As far as war stories go, I've only actually used an older-model ISCAN
eyetracker, so this stuff might not apply to other models, but here
are a few things we learned:

- Eyeglasses will probably need to be removed, and maybe contacts as
well. Most eyetrackers work by observing light reflections on your
eyes, and eyeglasses/contacts tend to mess those up.

- Mascara is evil. In our trials mascara tended to add a LOT of noise
to the input. Eyeglasses/contacts fudged the input by a good bit, but
didn't add much noise or chaos; mascara added chaos.

- It might take several minutes to reset or recalibrate the software
between users or trials. We found that we needed to recalibrate the
eyetracker every 15 minutes or so (2-3 trials, for us), and this
altered our schedule somewhat. I don't think we ever found a final
reason for the gradual loss of calibration.

- Give the user some time to get acclimated to the lighting conditions
in your lab. We had a fairly dim lab setup (all the windows were
covered up, to maintain a constant level of brightness) and it would
take the user's eyes several minutes to get acclimated, especially if
it was bright and sunny outside. This time could be spent on an
introductory survey or something similar.


I hope this helps, and good luck. =)

----------------------------------------
Steven Pautz
design newbie