[lug] An alternative keyboard and its issues
riddochc at gmail.com
Mon Sep 22 19:14:57 MDT 2008
I've got a little problem I'm trying to solve, and could use some help.
I recently got myself a new and very bizarre device which will help
me type, and is great for my wrists, but lacks a certain something.
I'll get to that. First, how it's used. There are two joystick-like
things which slide in eight possible directions, like old arcade
joysticks. When you slide both to the left, it sends the keyboard
scancode for 'a' to the computer. It can be used as both a mouse and
a keyboard. Their documentation calls the eight possible directions
"north," "northeast," etc.
Here's the design problem I'm trying to work around: They simply
arranged the key positions sequentially. "a" is west on both the left
and right "orbs." "b" is northwest on the left, and west on the right.
"c" is north on the left, and west on the right. It's not an ideal
layout. But layouts are easily fixed with xmodmap... if only I knew
what to move them *to*.
This keyboard seems to be able to function just fine with me not
actually re-centering after moving to a key's position. As a result,
I could hold one orb in place while moving the other to a different
position - this would be less physical effort than moving both orbs on
every "keystroke." Optimizing for that alone would be an improvement,
but I can think of at least one more reasonable consideration: sliding
to either the left or right from the current position is easier than
having to aim for a position not quite across the orb -- or worse, a
quarter turn away.
So, given a set of 47 remappable key positions (some just aren't for
reasons too complicated to go into) I'm trying to find a mechanism to
decide on where each key *should* go. I have statistics on which
letters follow which others with what frequency. I can get statistics
on the timing between different orb movements. Which brings me to the
last part: by what means would I use this information to find
(reasonably) optimal positioning for the keys? This is just a bit
Some of you may remember a guy who wrote an evolutionary algorithm for
trying to optimize his keyboard layout. You'll notice that I'm
trying to use a similar approach, but I'm having trouble getting from
here to there. This is clearly similar, but the differences are
enough that I'm getting stuck.
If anybody's interested in trying out one of these keyboards, I'll
bring it to the HS meeting at Cafe Sole this Thursday.
 The Orbitouch: http://www.keybowl.com/
More information about the LUG