From: salmo@elsie.cogsci.uwo.ca (Brian Fisher)
Subject: Re: TECH: Eye tracking systems?
Date: Tue, 20 Jul 1993 14:09:36 GMT
Organization: Centre for Cognitive Science



In article <2282le$oiu@news.u.washington.edu>, burke@ira.uka.de
(Wolfgang Burke) writes:

|> I would like to know more about eye tracking systems. What approaches
|> to solve the problem exist, how they work, who is doing research, who
|> builds such systems etc.?

I'll also respond here in svw, since these requests seem to occur on a
semi-regular basis. I've been responding by e-mail to requests for
eye monitoring information every now and again. Perhaps some of this
ought to be in a FAQ?

First, a disclaimer: I'm a research scientist working on the way
information is integrated between the senses in virtual environments.
I've been doing eye monitoring for ~ 10 yrs.  Eye monitoring is always
a challenge, especially in XY. The following are my opinions _only_
about the systems that I am aware of. I have used a variety of IR
systems, but not any of the others- hence the differences in level of
detail in my descriptions. Of course since I do interact with
colleagues that use the various systems, so I can relay their opinions
as they have been presented to me.

Common methods are: 

1) Search coil- measures the spatial relationship between a contact lens-
   imbedded coil of wire and an external set of coils placed around the
   subject's head. This requires the user to wear a contact lens with wires
   hanging off of it. Not a pleasant experience, and distracting at best.
   generally you can only stand it for 20 mins. or so. But this is the most
   accurate system.

2) EOG- uses DC amps recording from the extra-ocular muscles. Good for 
   monitoring fixations but not really an eyetracker. And the electric fields
   will play havoc with any magnetic headtracker. 

3) Perkinje image eyetracker (SRI)- The Cadillac eye tracker. Uses a set of 
   mirrors and servo-motors to align specularities (highlights) produced by 
   the reflection of an IR beam off of the lens and cornea of the eye. 
   These are very accurate, but quite expensive. They have had a reputation
   of being delicate (they tend to break down). SRI says that newer models 
   are more reliable. They are best for small excursion saccades-
   big saccades may exceed the capacity of the servos to track. This problem
   may also have been addressed by SRI. It is a big unit, and so it might be 
   tricky (impossible?) to integrate an SRI with a helmet-mounted display. 
   Has anyone done this?

4) IR scleral reflection systems- These use an IR camera or set of photocells
   to search for the image of the eye (cornea/pupil against the sclera). Most
   are analog systems that will require an A/D converter to be useful. Data    
   Translation and Metrabyte are common A/D converter suppliers. They'll run 
   you about 1000 US. 

  A) Video systems: There are a bunch of these, like the one from the U. of
     Toronto folks (reingold@psych.toronto.edu, fax: (416) 760-0675), or the 
     ISCAN model (fax: 617-868-9231). They are reasonably accurate but slow- 
     a sampling rate of 60Hz is common, but up to 240 Hz. is possible. 
     Processing time is extra, though, so you may have upwards of 50msec. lag
     for the 60Hz. machines. They are probably the easiest systems to use.
   
  B) Photocell systems: These are harder to use and more prone to
nonlinearities, but are faster- Most run comfortably at 1kHz. The only
XY systems that I know of are:

     1. Dr. Bouis, an IR scleral reflection system. A monocular verson for 
        human 
        subjects was 23,100 DM. You can get a Binocular version for 42,945 DM.
        Their fax is 49-721-45809 (as of 1991, when we bought ours). These 
        will 
        require some non-trivial software for calibration. We use this system.
        They make a helmet mounted model with a headtracker as well.

     2. There is a new system that is called "Ober 2" made by a place in Sweden
        They are also XY scleral reflection of IR light. Their recording device
        fits on the eyes like a pair of glasses, so should interfere less with
        helmet mounted displays than some. Their US dealer is Permobil Inc. 
        (fax 617-229-9841).

Good luck!

Brian D. Fisher, Ph.D.
Research Scientist
Centre for Cognitive Science/Institute for Robotics and Intelligent Systems 
University of Western Ontario, London, Ontario Canada