Path: utzoo!utgpu!news-server.csri.toronto.edu!rutgers!apple!usc!cs.utexas.edu!ut-emx!ccwf.cc.utexas.edu
From: awessels@ccwf.cc.utexas.edu (Allen Wessels)
Newsgroups: comp.human-factors
Subject: Re: Eye Movement Tracker
Message-ID: <50591@ut-emx.uucp>
Date: 16 Jun 91 02:32:46 GMT
References: <164661@felix.UUCP> <1991Jun14.192945.16681@linus.mitre.org>
Sender: news@ut-emx.uucp
Reply-To: awessels@ccwf.cc.utexas.edu (Allen Wessels)
Distribution: na
Organization: The University of Texas at Austin
Lines: 45

In article <1991Jun14.192945.16681@linus.mitre.org> marsh@thelonius.mitre.org (Ralph Marshall 617 271-8784) writes:

>The problem is only partly with the fact that existing devices for
>tracking your eyes are cumbersome and unpleasant to use.  There are
>reasonable systems that track your eyes visually using cameras and
>simple image recognition, but they still require that you keep your
>head in a 1' cubic area.

If the restriction of range is with respect to the cameras, you could make
some sort of headgear that would carry some of the sensors/cameras.  I don't
know what the theoretical limitations would be, but I'd think with fiber
optics you could do something.  There was an unsuccessful attempt at a 
sonic pointer via headgear marketed for the Macintosh a few years back.  It
wasn't uncomfortable to use, and some people could adapt to it fairly
quickly.

>The bigger problem is that you don't really look at things the way you
>think you do.  First of all, blinking can be a problem, but bigger
>problems come from the fact that your eye doesn't stay fixed on any
>one place.  It moves about very rapidly, and only generally stays in
>the area that you think you're looking directly at.  It also jumps
>when you move to look at another area, and takes a fairly long period
>of time to settle down at the new position.

It's been a few years since I played with an eye-tracker in a lab, but I dont
remember any serious problems with losing tracking due to blinking. The
question is whether the device could be usable in spite of the microsaccades
and saccades (jumps in fixation).  You don't have to get the machine to
do all the work.  As long as the error in the system is less than the
precision you need, people can adapt.

>So, you can't just replace the current pointing device with an eye
>tracker, even if it could track your eye perfectly.  You need a new
>model of how to determine the "selected area" based on the raw eye
>position values, and this work needs more research into how people
>really see things.

My intuition is that you could use an eye tracker as a mouse within the
current limits of technology (perhaps not economically.)  It is hard to
keep the eye fixated very long, but all you need is the duration of a
blink, or perhaps two.  One blink to indicate the beginning of a selection,
and two for the end.

All the equipment exists.  Someone just needs to decide that the experiment
is worth the resources necessary.