[misc.handicap] CBFB_REVIEWS scanner.txt

robertj@tekgen.bv.tek.com (Robert Jaquiss) (05/13/91)

Index Number: 15533

                      SEARCHING FOR THE AFFORDABLE SCANNER

                                  By Al Kuschel

Are you tired of having to wait for someone with sight to come and help you
with your mail, pay your bills and so on?  Or are you tired of sharing your
financial and personal business with others who have sight?  Now there's an
affordable solution to these problems.  That is an optical scanner with
intelligent OCR software.

An optical scanner is a piece of hardware that you can hook up to your
computer that will scan printed material into a graphics file and enter it
into your computer's memory.  With the OCR (Optical Character Recognition)
software, the graphics file is converted into a text file and then you can
use your computer's speech to read the text file.  Thereby converting your
computer into a reading machine.  You can also save these files in several
forms as plain ASC II text files or as your favorite word processing text
file.  In this manner, you can not only read books, magazines, manuals and
printed mail, but you can also save them in files to review later on.

Currently, there are many scanners on the market.  Most of them are for use
by sighted people and a couple were developed especially for blind users
such as the Arkenston and TrueScan.  Since I first heard of the Arkenston
Reader, I was interested in optical character recognition (OCR).  This is
the process which one can scan printed material into an ASC II text file,
and if you have a talking computer, the computer can then read the printed
material to you.  The reason I was so interested in the Arkenston is that
it was reported to be able to read printed material monospace and
proportional, typewritten material as well as dot matrix material which
just about covers everything from reports, books, manuals and magazines as
well as printed mail.

Immediately, thoughts of being able to read my own mail, textbooks, manuals
and so on rushed through my head.  Then, the stark realization of the
price--about $4,000.  I simply could not afford that amount.  So I set out
on a search for an affordable scanner that would do the job.

Originally, scanners were developed for government agencies and large
corporations and operated on closed systems.  This means that a scanner
could read printed material that came from this system.  if you used
printed material from a different system, chances are the scanner would be
unable to read the printed material.

The reason the goverment agencies and large corporations were interested in
scanners is that this would save them keying-in time and reduce
typographical errors.  The problem with these systems is that they were
able to work on their own closed system and were very expensive.  That is
why only government agencies and large corporations had them.

Then scanner companies began to develop trainable OCR software.  That is,
people could scan printed material into the computer and then actually tell
the OCR software what characters it was looking at and thereby train it on
different fonts so that the OCR software would then be able to read more
fonts and more styles of print.  This made the use of optical scanners more
widespread and with the increase in demand and mass production, the price
of optical scanners came down.  The problem with this is that it is
time-consuming to build up a library of fonts that the OCR software could
decipher.  But once you had your library built, practically any printed
material could be distinguished by the OCR software.

Then came the introduction of laser printers which were able to produce
print in a huge number of fonts and styles which then created the necessity
to retrain the OCR software for each different font style which meant that
companies were spending more time retraining the OCR software than they
would if a typist just keyed in the new material.

Then scanner companies developed intelligent OCR software.  That is,
software that could distinguish characters by their shape rather than the
exact copy of a letter so that different fonts of different sizes and
styles could be distinguished by the OCR software.  This enabled more
businesses to take advantage of scanning data into the computer rather than
keying it in.  Generally, a desk-top model used by a sighted operator would
cost from $5,000 to $8,000.

Enough of the history of optical scanners.  As I set out to fulfill my
dream of finding an affordable scanner that would enable me to read my
mail, textbooks and whatever else I would care to read, I found that the
flatbed and desk-top model scanners with automatic sheet feeders cost
approximately $2,500 without OCR software, and there was no guarantee that
they would work with speech.  My objective was to find a scanner that would
perform similar to the Arkenston Reader for less than $1,000.

This may seem to be an impossible task since the scanners themselves cost
about $2,500.  However, there is another class of scanner called a hand
scanner.  A hand scanner has a scanning window width of from two and a half
to eight and a half inches and as its name implies, it is powered by your
hand.  That is, you place the hand scanner at the top of the page and
slowly roll it down the page, scanning a strip of the page approximately
the width of the scan window.  As you can see, to scan an eight-inch page,
it may take more than one pass to scan the complete page.  The software
that comes with these scanners would then piece the scanned page back
together in the computer's memory.  Then, with the use of some type of OCR
software, you could convert the scanned image which is a graphic file into
an ASC II text file.

The first type of hand scanner I tried was a two and a half inch window,
trainable OCR scanner, which meant that once the scanned image was in the
computer's memory, the software worked similar to a spell-checker.  That
is, the curser would fall on a character and you would train the OCR
software by pressing the key on your keyboard corresponding to the
character on the screen.  Once the OCR was trained in this fashion for all
the characters in the alphabet, it could read that font and style from then
on.  This was encouraging to me since the price of this type of scanner was
from $300 to $500, depending on the OCR software and the width of the scan
window.  The biggest drawback to this is the fact that you had to train it
to read every character of your printed material.  That is, every size and
style.  Also I had much difficulty moving the scanner accurately down the
page in the middle where there were no refferance points.

The next type of scanner I tried was a hand-held scanner with a window
width of about five inches with OMNIFONT Character Recognition and
trainable.  What OMNIFONT Character Recognition means is that the software
can distinguish letters of all different font sizes.  That is, font sizes
between say 6 points and 24 points.  What that means is a print letter to
be small or very large.  This software distinguishs a large number of
characters but again, you would have to train it to distinguish the
characters that it could not recognize.  Again, there was the problem of
scanning down a page and having enough overlap of the scanned strip so that
the software could piece the scanned image back together again.

I found this scanner to be almost usable.  A blind person could use an
Opticon to train the software to distinguish the characters it could not
recognize.  By this method, you could train the  Character
Recognition to read just about everything.  The price for this type of
scanner and the  OCR was under $950.  Still below the $1,000 range
I was looking for.  But I did not have an Opticon to use to train the OCR
software.

The next scanner I tried was a hand-held scanner with a scan window width
of eight and one-half inches which I liked because that meant that one pass
would generally cover the whole page.  It also had a ten-page sheet-feeder
cradle attachment.  With the scanner placed in the sheet-feeder attachment,
you could stack up to ten sheets of printed material and it will scan the
sheets automatically one after the other.  The scanner's name is the Dest
Personal Scan scanner and can be used as a hand scanner for bound books or
a sheet-feed scanner for single sheets.  It also came with intelligent OCR
software called Recognize which can distinguish characters from 8-point to
18-point, the usual font sizes for the home, school and office and almost
every style of print.  It will read printed material, monospace and
proportional, typewritten material and dot matrix material.  The software
also contains a keyboard translater which will enable you to enter your
scanned documents into a word processor with all the commands of that
particular word processor to do underlinging, centering, setting of
margins, indenting, blotface printing, italics and so on.  From a menu, you
can choose from word processors such as Word Perfedt, Display-write 4,
Display-write 3, Multimate, Wordstar and many others.  The software can
distinquish print that is italics, boldface and underlined as well as
characters that touch each other since it uses the shape rather than an
exact match of a letter.  However, the better the print copy (such as
lightness and darkness and clear characters and non-touching characters),
the better the conversion from a graphics file to a text file.

Other features of the scanner are that when using it as a hand scanner to
scan bound books and landscape pages (pages that are printed across the
length of the page rather than across the width) and other documents that
won't fit into the sheet-feeder, you have to move the scanner across the
page slowly.  If you hear a low-pitched tone, that means you are moving too
rapidly and should slow down your movement.  If you hear a high-pitched
tone, that means that you moved too fast and have lost data and should
abort the scanning and rescan.

Another feature is that the software automatically eliminates graphics.
This means that you do not have to block out graphic material on a page in
order to scan.  I have seen demonstrations of the Arkenston and TrueScan
that will try to decipher graphics and after five or ten minutes will
decide that it is graphics and cannot read it.  This will not happen with
the Dest Personal Scan scanner since it recognizes the fact that it is
graphics and automatically skips it.

The program works well with speech on an AT compatible computer or 386
computer with at least 640 K of RAM.  It is able to scan a multip-page
document into one continuous file which you can import into your word
processor with all of the formatting commands in place so it will save you
lots of keying-in time.  The retail price of the scanner is $995.

I was so impressed with this scanner that I became a dealer.  Anyone
interested in this product can see my ad elsewhere in this issue.