Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.1 6/24/83; site fortune.UUCP
Path: utzoo!linus!security!genrad!grkermit!masscomp!clyde!ihnp4!fortune!rpw3
From: rpw3@fortune.UUCP
Newsgroups: net.cog-eng
Subject: Re: mice, pens, and graphics - (nf)
Message-ID: <2048@fortune.UUCP>
Date: Mon, 19-Dec-83 04:02:35 EST
Article-I.D.: fortune.2048
Posted: Mon Dec 19 04:02:35 1983
Date-Received: Tue, 20-Dec-83 06:42:12 EST
Sender: notes@fortune.UUCP
Organization: Fortune Systems, Redwood City, CA
Lines: 46

#R:umcp-cs:-429100:fortune:29300002:000:2185
fortune!rpw3    Dec 19 00:36:00 1983

John DeTreville's response about transformations to mouse input brings
something to mind that a trackball/joystick salesman said to me at
COMDEX nearly two years ago. He was complaining that nobody gave
trackballs a fair chance because nobody used them right. His points
were (as best I can recall):

1. Mice are good at pointing because humans have a predominately
   translation/position-sensitive feedback mechanism in performing
   pointing actions. Mice supply translation-style input; balls,
   joysticks, thumbwheels (Tek 4010), and dials (HP 9816) don't.
   (Light pens supply the right input, but the resolution stinks.)

2. BUT... If you pre-transform the input data from the acquisition
   device such that the transformed data meets the expectation
   of the human kinesthetic servo-loop, you can get mouse-equivalent
   or (here's the kicker) better performance out of some of those
   others.

His claim was that) they had hired a consulting firm to do the analysis
of the servo-loop parameters (impulse response, poles, zeros, gain, etc.)
[loop = eye -> brain -> hand -> pointing device -> software -> display -> eye]
for a particular (military) target acquisition task, and that they had
found that a force-operated joystick (one that doesn't move, you just
push on it) was as good [good = f(speed,accuracy,fatigue)] as any of
the devices listed above. In fact, even the thumb-sized ones on the end
of a control stick could be made nearly as good.

The computations weren't particularly complex, and could have a fairly
low accuracy, as long as [here's where John DeT's note reminded me]
zero-input => zero-motion and small-input => small-motion. (Also,
the dead-zone should not be much (if any) bigger than a small-input.)

Obviously, the salesman was trying to sell me his product, which was
trackballs and (mostly) force joysticks. But some of his devices were
MUCH cheaper than mice, so I'm curious... Does anyone else know
of more work on "pre-compensating" human input with "matched filters"?

Rob Warnock

UUCP:	{sri-unix,amd70,hpda,harpo,ihnp4,allegra}!fortune!rpw3
DDD:	(415)595-8444
USPS:	Fortune Systems Corp, 101 Twin Dolphins Drive, Redwood City, CA 94065