Speech For The Deaf

I am linus's ho writes "CNN is running a story about gloves which transelate sign language into audble speach, in a stephen hawking type mannor, only, i suppose, much different. The article can be found here"

Comments

I am linus' karma ho

[philam3nt]

A much more informative page that explains the technology (called GRASP - Glove-based Recognition of Auslan using Simple Processing) briefed in the CNN article can be found on Waleed Kadous's website:

GRASP Site:
http://www.cse.unsw.edu.au/~waleed/thesis.htm l

More generalized Gesture & Sign Language Recognition Research:
http://www.cse.unsw.edu.au/~waleed/gsl-rec/

Also see the self-proclaimed Gesture Recognition Home Page (good resource, tons of links)
http://www.cybernet.com/~ccohen/

Or just search google like I did for 'Machine Gesture Sign Language' and get a wealth of links.
</whoring>

Re:Why? (researcher's response)

[waleedk]

Firstly, there's an even simpler solution than a keyboard: a piece of paper and a pen. Most (non-geeks) can write faster than they type anyway, so why bother with the keyboard at all?

Secondly, I feel it does offer an advantage -- it uses the medium of communication that the Deaf themselves prefer to use. Sign languages are not a word-for-word translation of English. No offence to Deaf people intended, but if you've ever read any unedited text by Deaf (say, e-mail), their grammar is sometimes poor. It's not because Deaf are stupid, it's just it's not their language, and the language that they do use -- ASL or Auslan or whichever sign language -- is so totally foreign to spoken language that it is hard for them to come to grips with the grammar. So forcing them to use English is already forcing them to do something they'd prefer not to do. It's like saying why bother with Babelfish when anyone could learn Spanish?

Re:Isn't this old? (researcher's response)

[waleedk]

Actually, I did a bit more research on Ryan's work, and it turns out what he built was a system that converted fingerspelling, not sign language into speech.

Sign languages typically have a finger-spelling system as a fallback. Say you want to specify a name, like McGill. There's no sign for it, so you finger-spell it. Each letter has a corresponding sign. ASL has single-handed fingerspelling, while Auslan has two-handed fingerspelling.

But it's not sign language. A typical sign might be something like "thank" where you touch the chin and move the hand forward (at least in Auslan anyway).

Recognising sign is much harder. Fingerspelling is pretty much position-independent. For sign recognition, you need to track the person's position and motion.

This is not to cast any aspersions on Ryan's work, of course -- especially that he made it so cheaply. I think that's a major accomplishment. But designs for such gloves date back to Grimes' work in 1983 (patent 4,414,537), or James Kramer's work at Stanford in the early 90's (patent 5,047,952).