[sugar] Image Recognition

[Kevin Cole]

On 3/18/08, Benjamin M. Schwartz <bmschwar at fas.harvard.edu> wrote:

> Higher rates of deafness in the OLPC target countries make this even more
>  important.  The XO has a distinct focus on real-time text-based chat
>  ("IM").  I suspect that this will do a lot to improve communication
>  between deaf children and hearing children, much as it seems to have done
>  in the US.  Chat also provides an easy opportunity to bring deaf children
>  into the classroom, simply by designating one or two hearing students as
>  transcribers for each lesson, and letting the trancribers and deaf
>  children share a chat session.

While it's true that IM and closed-captioned television have been a
real boon, many deaf
students have very poor written language skills -- particularly those
who were born deaf or lost their hearing before age two. The term used
is "prelingual deaf".  Unless prelingually deaf children are immersed
in sign language from birth, the cognitive system that processes
language quite often does not form correctly or completely.
Nonetheless, I'd agree: Chat can be very helpful for ALL deaf users.

>  Regarding sign language, I doubt that automated video analysis and
>  transcription of sign languages will be possible.   The CPU is similar in
>  processing power to the best desktops of 1997.  In 2000, real time video
>  analysis could recognize fewer than 100 gestures, with an error rate of
>  almost 10% (http://doi.ieeecomputersociety.org/10.1109/ICPR.2000.906112).
>  ~ Today, sign language recognition is still just a research project, even
>  with unlimited processing time.
>
>  Video capture is already possible on the XO, and may be useful for
>  learning sign language if the teacher is only available over the internet.
>  ~ Video playback could certainly be useful for learning a sign language,
>  perhaps with a video course provided on a USB stick.

I was thinking more along the lines of the video capture rather than
analysis and transcription. There are also some efforts underway to
produce clear animated sign a la high-quality Second Life actors.
I've seen one demo that I was wowed by.  I've got the URL filed away
somewhere...

>  There appear to have been a number of efforts to codify written forms for
>  sign language, notably "SignWriting" (signwriting.org).  There is even a
>  Python-based program for "SignWriting" (http://signwriter.takdoc.de/).  Is
>  this writing system, or another writing system, viewed favorably by the
>  community at Gallaudet?  If so, creating a sign-language writing Activity
>  for OLPC might be an appropriate endeavor for Gallaudet students.

Honestly, I don't know for certain.  There was a newspaper for deaf
readers with a wide circulation -- "Silent News" if memory serves.
For a time, it offered a back page which was all in a sign writing
system, but I believe it received very little interest from anyone,
and was dropped. I don't think people are particularly hostile to it,
but view it as a bit "silly".  I've been at Gallaudet a long time and
cannot think of anyone who took it very seriously -- other than a few
academics who were convinced they could get an article or grant money
out of it. ;-)

-- 
 Kevin Cole                    |  Key ID: 0xE6F332C7   (GPG/PGP)
 Gallaudet University          |     WWW: http://gri.gallaudet.edu/~kjcole/
 Hall Memorial Bldg  S-419     |   V/TTY: (202) 651-5135
 Washington, D.C.  20002-3695  |     FAX: (202) 651-5746

           ". ! 1 |" -- Rene Magritte's computer