From: dstamp@watserv1.waterloo.edu (Dave Stampe-Psy+Eng)
Subject: Re: Integrated laser arrays for Eyephones
Date: Wed, 6 Nov 1991 05:05:41 GMT
Message-ID: <1991Nov6.050541.29147@watserv1.waterloo.edu>
Organization: University of Waterloo



70353.3056@CompuServe.COM (Christopher Fry) writes:

>>I'm not aware of any current research into eye-tracking displays: who's 
>>doing this?  
>
>Iscan Inc, 125 Cambridge Park Drive, Cambridge, MA, USA 02138
>(617) 868-5353
>
>I saw these guys at a trade show about 2 years ago. They had 2 models:
>a desktop mounted mondel that was a video camera sitting at the base of your 
>comupter screen looking up at your face. It might have had motors in it to 
>track your head. It output the position of your pupils. As I recall, their
>demo was to highlight text that you looked at, and scroll it as need be.
>Their second model was mounted like the rear-view mirror clipped onto glasses 
>frames that's used by bicyclists. It contained a tiny TV camera. 

I've worked with Iscan's equipment in the past: in fact, their pupil-
tracking computer card is part of my latest gaze-tracking system.  It
costs about $10,000 (but could be duplicated for a LOT less, judging
from my experience with it).  What I meant was whether anyone is 
currently using the graded-resolution display which responds to eye
position that I described.  All I know of is the 20-year-old attempt
by McDonnel Aircraft.

>>Head mounted displays, to my mind DO NOT qualify as a good
>>way to reduce pixel counts, even with high-resolution areas (see below).
>
>Isn't the ideal to have an eye-tracking display mounted on the head?
>Big movements handled by your head motions. Little movements by something 
>small and light. 
>
>The fast-light, 2-d tilt mirrors developed for aiming the lasers in 
>laser-light shows may find some application here. If we can tolerate delays as
>long as 100MS [Stampe's numbers, sounds long to me but I haven't done the 
>experiments] then the laser aiming mirrors are plenty fast enough. Probably 
>not quite light enough, though.

Actually, you have to be prepared to track BOTH large and small motions
with the eye-following displays.  Experiments that tracked both eye and
head motions show that subjects tend to rotate the eye over 30 degrees,
then follow up with a head movement in response to stimuli presented
off to the side.  For a "transparent" VR vision system, we must not
try to change the user's natural use of head or eye movements.

The 100 mS numbers are the McDonnell Aircraft results: I feel better with
about 50 mS myself, based on experience with eye-trackers (that is,
devices which MEASURE eye movements, not ones that follow them).  Sometime
in the next year, I hope to modify the current software to change a 
displayed screen image based on gaze position: this could answer a few
of the questions.  I feel that a long delay will be OK for large eye
motions, but not for eye movements of less than 5 degrees.

>>Here are some numbers on the advantages of eye-tracking displays over
>>the standard head-mounted eyephones: [...]
>
>Bravo! You get a standing ovation from me for this message! I've heard 
>unsustantiated talk about cutting down needed pixels by decreasing res from 
>the fovea out, but had no idea that the savings were potentially so dramatic.
>100K -> 150 K pixels is no more than current LCD's can put out. With 
>"distorting" lenses we can leave them in the regular array as they're 
>manufactured for TV yet present the eye with varying resolution. But those 
>LCD's are TOO BIG to move quickly with low weight motorsso we're back to some 
>kind of integerated-optics light sources, be they LED's or microlasers. Only 
>now we need a LOT less than my original proposal.
>
>Once we've got the display itself, a bunch of clever graphics software needs 
>to be written. I expect each pixels to take considerably longer to draw than 
>regular-array square pixels [and remember, we need to fill an oval, not a 
>rectangle].  If our general-purpose hardware still isn't fast enough for 
>manipulating those 150K pixels, then this sounds like an ideal candidate for 
>special purpose parallel hardware. 

There are other options than distorting lenses, of course.  If we DID have
our 32 Mpixel display, we could distribute the 150K pixels via hardware
to simulate the graded resolution.  But the lens idea sounds a lot simpler.
Actually, the pixels end up wedge-shaped toward the periphery of the display.

Mirrors would probably not give a big enough FOV, but there are a LOT
of hybrid optical solutions available.  I've thought of a number of ideas,
including magnifying mirrors which reduce the display and lens optical
angle, and reduce display swiveling requirements to a small angle. 

--------------------------------------------------------------------------
| My life is Hardware,                    |                              | 
| my destiny is Software,                 |         Dave Stampe          |
| my CPU is Wetware...                    |                              | 
| Anybody got a SDB I can borrow?         | dstamp@watserv1.uwaterloo.ca |
__________________________________________________________________________