On Tue, May 09, 2006 at 11:28:23AM -0000, Eyetracking@yahoogroups.com wrote:
>[..]
>As we are novices in the field of eye tracking – we started to use
>eye tracking (also the Tobii system) six months before – we are very
>interested to learn from others' experiences. We use eye tracking
>for website testing as well as for dynamic video stimuli (e.g. in
>our iTV project). I read in the workshop slides, that dynamic
>stimuli are one of the most challenging areas. I'm relieved to hear
>that not only we had some problems to manage with this specific
>context.
>
>I've also some comments / questions related to the workshop slides
>and to your work:
>
>- During the workshop, were there reported some practical
>examples/experiences about the use of eye tracking for dynamic
>stimuli? (some links, interesting contacts…)
>
>- Did you ever manage to define AOIs for dynamic stimulis with the
>Tobii system?
>
>- How do you combare eye tracking data from different users from the
>same usability test?
>
>[..]

To handle dynamic stimuli, one idea (current best practice?) that was
suggested was to re-instrument whatever application you had to keep
track of dynamic AOIs (or objects) and eye movements. This usually
requires adding code to the application of interest, which is most
likely a significant undertaking. Still, the Tobii eye tracker does
provide an SDK to allow this (I believe using C++ or C# under Windows;
there is also a Linux client that we wroet but you'd need to ask Tobii
for permission in order for me to release it).

We at Clemson have been doing this kind of thing almost exclusively
since we generally develop our applications from scratch (in the
Linux environment). The downside to this is that our applications
are not what users typically use and are academic creations lacking
the sophistication of commercial products. However, for basic research
problems, these types of apps often suffice in elucidating a studied
phenomenon.

One of our most recent practical examples comes from students in my
eye tracking class last (fall) semester, where they built a low
fidelity driving simulator. An example of a moving AOI was a
hazardous vehicle in a 3D dynamic driving scene. To be more
technically specific, the AOI was a bounding box around the vehicle,
if I remember their implementation correctly. The simulator resembled
a first-person driving game, where the hazardous vehicle would move
relative to the viewer (e.g., pass at high velocity, or swerve ahead
of the driver). They should be in Sarasota, FL tomorrow presenting
their results. If you can't make it there, a preprint of their poster
is available here:
<http://andrewd.ces.clemson.edu/research/vislab/docs/VSS06.pdf>

From the poster you should at least get a feel for the application
and the coding effort that was required to track the moving vehicles
and count fixations upon them. In 3D parlance, what is going on
is unprojection of the (x,y) gaze cooridnate into a ray shooting
through the (x,y) screen pixel into the screen, then a simple
ray-casting intersection test with the 3D dynamic stimuli visible
in the view frustum. In 2D applications, you wouldn't need to do
the 3D back projection (hence just a simple point-in-polygon query),
but would still need to track where the AOI was at each "frame" of
the simulation (or instance in time) to allow synchronization
(intersection test) with the eye movement sample at time t.

>Sorry to ask so many questions, but I'm really curious to learn from
>your experience. If you like to know more about our research areas,
>please visit: http://www.icts.sbg.ac.at . For information about our
>attemps to integrate eye tracking into traditional usability tests,
>please feel free to contact me - even if we are just starting to use
>it :)

Those are all interesting projects although I'm not sure in which
one you're attempting to use the Tobii -- iTV?

Cheers,
Andrew

--
Andrew T. Duchowski, Assoc.Prof. 309 McAdams Hall (area code: 864)
Computer Science Dept., MS-0974 tel:656.7677 fax:656.0145 cel:247.2050
Clemson Univ., Clemson, SC 29634 duchowski@...
http://www.e-t-r-a.org http://andrewd.ces.clemson.edu