The number of vectors needed to represent handwriting is huge. It would
help to use something like NURBS to reduce the data size, but that requires
yet another transformation. If you use a raster "canvas", then you need to
map the mouse movements to the canvas. If the resolution of the canvas is
low, the data is smaller, but you have a loss of the small details. Using a
higher resolution canvas captures the details, but greatly increases the
data size.
This is why such applications are usually run locally. You have all the
horsepower available to transform the handwriting into data. But Scott has
a set of goals that make this approach undesirable to him.
Like I said - this could be difficult. But I understand the attraction of
the challenge.
Ken Elliott
=====================
-----Original Message-----
From: slug@nks.net [mailto:slug@nks.net] On Behalf Of Setheck
Sent: Monday, June 22, 2009 6:17 PM
To: slug@nks.net
Subject: Re: [SLUG] Open Source handwriting recognition application?
I understand an input device sends a stream of information, but why
would you want the stream, instead, wouldn't you just want to use a
web drawing canvas type app? If you could do it with vector graphics
you would lose the problem with resolution. But that may open another
can of worms.
On Jun 22, 2009, at 2:54 PM, Ken Elliott wrote:
>>> " I'm confused by your response. If you map an input device like the
> tablet to the mouse and use ajax to capture only mouse input from the
> client. Wouldn't it be easy to store all serverside?"
>
> Pointing devices either send a stream of x/y deltas, or absolute x/y
> position. So the resolution will dictate how much data there is to
> process.
> It could be quite a lot. If you package that data as XML, it is
> quite bulky
> (as opposed to binary data). Now the server has to take that data,
> decode
> it and figure out what was written. All that just to get a few
> pieces of
> text. Add 1000 client and you have imposed quite a load to the
> server.
>
> Add to that the fact the a mouse doesn't do a good job with
> handwriting and
> you have a system that will have a difficult time satisfying the user.
>
> So if we solve the mouse problem by using a Wacom tablet, we now
> impose a
> limitation to the system. If this is a private system, where we can
> dictate
> hardware, then we can easily dictate the use of a local app. Use of
> a Wacom
> tablet and a local app means 1000 users will have the power of 1000
> PCs,
> each dedicated to resolving the handwriting-to-text transformation.
> This
> scales quite well and simplifies the server side.
>
> Is any of this a problem? That is why I asked what the input
> devices would
> be and other questions. Without that data, we will all be guessing.
>
> Ken Elliott
> =====================
>
> -----------------------------------------------------------------------
> This list is provided as an unmoderated internet service by Networked
> Knowledge Systems (NKS). Views and opinions expressed in messages
> posted are those of the author and do not necessarily reflect the
> official policy or position of NKS or any of its employees.
-----------------------------------------------------------------------
This list is provided as an unmoderated internet service by Networked
Knowledge Systems (NKS). Views and opinions expressed in messages
posted are those of the author and do not necessarily reflect the
official policy or position of NKS or any of its employees.
-----------------------------------------------------------------------
This list is provided as an unmoderated internet service by Networked
Knowledge Systems (NKS). Views and opinions expressed in messages
posted are those of the author and do not necessarily reflect the
official policy or position of NKS or any of its employees.
This archive was generated by hypermail 2.1.3 : Fri Aug 01 2014 - 20:20:15 EDT