On Nov 28, 2017, at 9:33 PM, Robert Ringwald via Talk
<talk@lists.window-eyes.com> wrote:
Thanks David. Very interesting.
My one question is, if all I want to do with the KNFB program is decipher
scanned PDF files that are already in my WLM email, will the KNFB reader read
those internally with no camera or any other device needed?
I have Win 10 and WLM 12 on a del PC.
-----Original Message----- From: David via Talk
Sent: Tuesday, November 28, 2017 2:16 PM
To: Window-Eyes Discussion List
Cc: David
Subject: Re: Question About KNFB and camera.
PDF documents, or any other file formats on your computer, are
electronically stored information.
Cameras cannot read electronic documents. They are such constructed,
they need to "see" things in real, before they can do anything.
All electronic formats, be it music, video, documents or just any
blah-blah-blah, will have to be processed internally in the computer.
To best illustrate things, may I suggest you think of the camera as the
"eye" of the computer. Let's at the same time, tell the scanner to be
the other "eye" of the PC. The CPU, (or processor), the RAM and any
other electronic inside your computer's physical box - well, let's name
it the brain. Even so, the hard disk or SSD, which we will compare to
your "deep memory".
As you well know, your physical eyes cannot "look" inside the brain, and
perform anything from within your body. Rather, the eyes can feed the
brain with information, which your brain now can process.
Back to your query. You will need a camera to feed any written or
physically visible information into the computer. Whatever has already
been fed into the computer, like an electronically stored document, will
be non-interesting for the camera, scanner or any further feeding
equipment. All processing of what you have in your brain, will be done
by the brain directly. All information already stored on your computer,
will be processed directly by the computer, and loaded software.
I know, you wanted a quick answer to your question. I just thought it
might be helpful for you and others, to have a clarified comprehension
of why the answer is the way it stands.
To jhust elaborate a tiny bit here, let me in very short terms tell you
how any OCR software works.
First of all, it needs to retrieve some information. It will typically
leave you the chance of defining whether it should grab some electronic
document, or if it should contact an external piece of equipment - like
a camera or a scanner. To the software, it basically does not matter
whichever way you feed it with information.
Next, it will start to process the information it has loaded into its
memory. All such electronic information is made up of 0's and 1's, also
known as pixels. And the software will compare the layout of these, with
an internal dictionary. The dictionary will be like a tremendous
collection of stencils. If the OCR finds that a set of dots (or pixels)
in the received information matches any stencil in the dictionary, it
will know what character this will represent. It now will "type" this
character into a virtual document, thereby imitating you pressing a key
on the keyboard.
Finally, when it has finished the whole loaded information, it will
present you with the virtually typed document.
For your information, in old times, the stencil-lookup was pretty much a
one-to-one comparison. That means, it would need a match that would be
very close to the exact stenciled shape. If it was to recognize anything
to be the letter O, it would need a set of pixels in a perfect circle.
Modern OCR software has become far mor "inteligent", whatever we want to
talk about inteligence when comes to silly electronic units like a
computer. The inteligence is that the OCR no longer will depend on close
to exact matches. To a very high degree, it might "look" at the
properties of a scanned character, and base its recognition on the
results thereof. For instance, it would conclude that a set of pixels
that resemble two parallel vertical lines, slightly spaced from each
other, with a horizontal line running just about mid-way up between the
verticals - all in all will be interpreted as the capitalized letter H.
Likewise, a vertical bar, with a tiny line pointing diagonally out from
the upper left end, will likely be told to be the number 1.
As you might understand, such propetary comparison will be more
forgiving, than if you were to compare exact matches. You no longer need
to define how high the character can be, or what the width should be.
The OCR can "see" this is the number 9, big or small print, simply by
recognizing the shape and other properties of the character. This is one
of the reasons, modern OCR can perform high degrees of faultless
recognition. In the old days of the 80's, often a number 9, and the
lower-case G, would be confusingly recognized as either, due to the fact
that they quite much would resemble similar pixel-patterns.
to improve the OCR recognition, modern OCR software further will hold
comprehensive dictionaries for spelling, in several languages. It is
considered very little likely, that any word in English would be:
log9ing,
so the OCR will recognize this as if it was a common typo, and replace
the 9 with a g, making the word:
logging,
which happens to be a validly spelled English word.
Since they now aday do propetary stencilized OCR, they also can perform
recognition of hand-writing. At least, to a certain degree. Simply by
attempting to recognize the shape and general makeup of the lines on the
paper, the OCR will conclude that your droddle "pretty much looks like"
a given character. By correcting the software whenever it performs the
wrong recognition, it eventually will "learn" the style of your
hand-writing. Such correction is what is known as
training the software.
Hope all of this was of any help and interest to you, or others.
David
On 11/28/2017 6:23 PM, Robert Ringwald via Talk wrote:
I do not have a smart phone. In order to use the KNFB program to just
read PDF's on the computer, do I need a camera? Or can it be done
internally within the computer?
Windows 10.
Bob Ringwald piano, Solo, Duo, Trio, Quartet, Quintet
Fulton Street Jazz Band (Dixieland/Swing)
916/ 806-9551
Check out my performing schedule: www.ringwald.com/schedule.php
Amateur (ham) Radio Station K6YBV
“If you don't read the newspaper you are uninformed,
if you do read the newspaper you are misinformed.” -- Mark Twain
---
This email has been checked for viruses by Avast antivirus software.
https://www.avast.com/antivirus
_______________________________________________
Any views or opinions presented in this email are solely those of the
author and do not necessarily represent those of Ai Squared.
For membership options, visit
http://lists.window-eyes.com/options.cgi/talk-window-eyes.com/trailerdavid%40hotmail.com.
For subscription options, visit
http://lists.window-eyes.com/listinfo.cgi/talk-window-eyes.com
List archives can be found at
http://lists.window-eyes.com/private.cgi/talk-window-eyes.com
_______________________________________________
Any views or opinions presented in this email are solely those of the author
and do not necessarily represent those of Ai Squared.
For membership options, visit
http://lists.window-eyes.com/options.cgi/talk-window-eyes.com/rs%40ringwald.com.
For subscription options, visit
http://lists.window-eyes.com/listinfo.cgi/talk-window-eyes.com
List archives can be found at
http://lists.window-eyes.com/private.cgi/talk-window-eyes.com
_______________________________________________
Any views or opinions presented in this email are solely those of the author
and do not necessarily represent those of Ai Squared.
For membership options, visit
http://lists.window-eyes.com/options.cgi/talk-window-eyes.com/jhud7789%40twc.com.
For subscription options, visit
http://lists.window-eyes.com/listinfo.cgi/talk-window-eyes.com
List archives can be found at
http://lists.window-eyes.com/private.cgi/talk-window-eyes.com
