nanogui: GPL Handwriting recognition demo under nano-X
Subject:
GPL Handwriting recognition demo under nano-X
From:
Gaillard Pierre-Olivier ####@####.####
Date:
13 Dec 1999 22:18:03 -0000
Message-Id: <38556962.6435FB41@free.fr>
Hello,
I have used the kanjipad GPL program to recognize roman characters as
define by Goldberg(93) under Nano-X.
This a very simplistic demo, not a real application. I hope some of you
will
be interested and help me test it on real hardware (I tested under X11
only)
and write a really usable keyboard.
Here is the README.nanox file that I have added :
This README file was added by Pierre-Olivier Gaillard
####@####.####
I have added a few files to the standard Kanjipad distribution but
refrained from changing the original jstroke files.
To compile the nano-X demo.
1) Edit Makefile to suit your nano-X install
2) make nano
3) demo.sh
4) draw in the left (text) and right (numerics) windows according
to newdoc/unistroke.fig. The detected characters will be printed
on the standard error with scores between parentheses (the lower
numbers are BETTER.)
What I did :
0) Mostly nothing, the algorithms are from Goldberg, Robert E.Wells and
Todd David Rudick. And Owen Taylor ####@####.#### is the maintainer
of
the kanjipad package. All the code is under GPL.
1) I used the unistroke alphabet to redefine the roman characters as in
textdata.h. numdata.h contains mappings of the 10 first letters of
the
alphabet to 10 figures (0-9).
2) I added documentation in newdoc.
- directions.fig : shows the way to code directions for jstroke
engine.
- unistroke.fig : show the definitions of 26 letters.
The original unistroke alphabet was defined by Goldberg in 1993 at
Xerox Parc. You can find references on the Web.
3) I wrote simple interface files :
- nanox_demo.c : the main nanoX program. Creates a window and
two subwindows. Accepts mouse input and prints characters detected
to standard error.
- dictionary.* : interface to the jstroke StrokeScorer.
does the printing to stderr.
- rawstrokelist.* : easy to use way to input strokes (we only
use one stroke characters, but it can handle more. I'd like to
work on Japanese input later on.)
What needs work :
1) define a good and efficient set of characters (2 actually, one for
the text window and another for the numerical window).
2) Write a better input program
3) Define and implement the interface between this program and the
rest of Nano-X.
Where to find updates :
http://po.gaillard.free.fr
I really would like some help. Especially advice regarding the
interface
with Nano-X. In fact it would nice if somebody who really knows
about user interfaces would show me how to do this well.
Thanks a lot.
P.O. Gaillard