nanogui: Thread: Handwriting recognition on handhelds.

I've been playing with writing a handwriting recognition program for Nano-x
and have had some success.

The code is far from complete but does function to the extent of recording
strokes and doing some simple translation of the 'point' sequence. When
finished the code will released GPL (it's just too messy at present to put
on general release).

This work is based on libstroke, scaling input character to 5 by 5 grid and
then outputs the sequence of 'cells'.

grid is:
	A	B	C	D	E
	F	G	H	I	J
	K	L	M	N	O
	P	Q	R	S	T
	U	V	W	X	Y

so AFKPU would represent a 'i' (or 'I' or '1' or '!' depending on mode)

I thought that this system would work best for a 'simplified' alphabet or
'graffiti' style characters. Any comments??

I'm now at the stage of recording the whole alphabet and I'm asking for
volunteers to offer time inputting characters in their own handwriting so
that I can get a range of styles..... email me directly if you're
interested.
Simon Wood

: The code is far from complete but does function to the extent of recording
: strokes and doing some simple translation of the 'point' sequence. When
: finished the code will released GPL (it's just too messy at present to put
: on general release).

Simon - this is something we definitely need.  However, it seems
that recognition isn't an easy process.  Is there any other open
technology we might be able to start from, so that you don't
have to reinvent all the recognition algorithms?

Regards,

Greg

----- Original Message -----
From: "Greg Haerr" ####@####.####
To: "Simon Wood" ####@####.#### ####@####.####
####@####.####
Sent: Monday, June 26, 2000 11:34 AM
Subject: Re: Handwriting recognition on handhelds.


> : The code is far from complete but does function to the extent of
recording
> : strokes and doing some simple translation of the 'point' sequence. When
> : finished the code will released GPL (it's just too messy at present to
put
> : on general release).
>
> Simon - this is something we definitely need.  However, it seems
> that recognition isn't an easy process.  Is there any other open
> technology we might be able to start from, so that you don't
> have to reinvent all the recognition algorithms?

libstroke, Squeak, and Itsy scribble are the only ones I know about.

Regards,
Brad

Brad said:
>Greg said:
>> Simon - this is something we definitely need.  However, it seems
>> that recognition isn't an easy process.  Is there any other open
>> technology we might be able to start from, so that you don't
>> have to reinvent all the recognition algorithms?
>
>libstroke, Squeak, and Itsy scribble are the only ones I know about.

Here's another -- just entering beta though.  AFAICS the scribble
recognition is statically built-in.

--j.

>From: Mike Bennett ####@####.####
>Date: Fri, 23 Jun 2000 11:42:05 +0100
>To: ####@####.####
>Subject: [ILUG] wayV: Gesture recognition as part of your desktop

Righto, for those of you who may be interested version 0.0.9 (first
public release) of wayV is now available at:
        http://wayv.sourceforge.net
in the download section:
	http://wayv.sourceforge.net/downloads.shtml

wayV is a free software project, which extends the functionality of the
mouse by allowing gestures to represent actions, i.e.  draw a C and xcalc
starts, draw an N and Netscape starts, etc.

It should be window manager independent and has been tested with Gnome,
KDE and WindowMaker.

Feedback would be appreciated :)

Mike    

Hello.

On Jun 26 we had a few messages about the subject matter.  I have a
few questions about this issue, as a few months ago I wanted to play
with the problem.

Can someone please give some pointer to current implementations? Are
they general algorithms or complete programs for microwindows or other
environment?

What about the patent of Xerox? Back then when I was pondering on
stroke recognition, I heard that Xerox sued 3Com for the pilot.
The patent is "US4034343: Optical character recognition system".
Is this an issue or not?

Alessandro Rubini wrote:
> 

> Can someone please give some pointer to current implementations? Are
> they general algorithms or complete programs for microwindows or other
> environment?
 - Squeak (Smalltalk) by Alan Kay seems to be very interesting because
this
guy knows algorithms from the 70s that CANNOT BE PATENTED (since they
were
published before the US patent office started delivering silly software
patents).
 - A C'T article presented some algorithm but I don't know what it is
worth 
  and lost the reference though you may find it back by searching
slashdot.
 - I modified a Japanese characters recognition program called kanjipad
to
  run under Nano X (and describe a roman character-set for it).
  Unfortunately that program is derived from the Xerox
  Parc article about Unistroke. So that it offers almost no protection
against
  a patent suit (we would win in the end, but what you are looking for
is an algorithm
  so old that the lawyers will leave you alone).
> 
> What about the patent of Xerox? Back then when I was pondering on
> stroke recognition, I heard that Xerox sued 3Com for the pilot.
> The patent is "US4034343: Optical character recognition system".
> Is this an issue or not?
> 
 This is considered as an issue by Alan Kay himself. Though, he is
convinced
  that there is plenty of prior art from the 1960s so that the patent is
void.
 The problem is : Who is going to fight Xerox about it ? Even 3Com might
 prefer to pay a license... And then Xerox can still use its bogus
patent against
 free software recognition programs (especially those that use
characters made of only one
 stroke like graffiti)...

 So, the issue is pretty difficult. I would go for Squeak myself. In
fact when I buy a handheld
 computer this autumn (probably the Ipaq from Compaq) I will evaluate
Smalltalk on it, use squeak
 or recode Squeak. As you can see I am not even considering using the
software (kanjipad) that I already have
 adapted : that is because you cannot make free software in the shadow
of a patent.


	Regards,


	P.O. Gaillard