This project at UCLA is such an intriguing idea. And using today’s hardware and software, not too hard to implement.
Sign of the times?
July 2nd, 2020
4 Comments
Got Something To Say:
Feeds
RSS feed© Copyright Quentin Stafford-Fraser
This project at UCLA is such an intriguing idea. And using today’s hardware and software, not too hard to implement.
© Copyright Quentin Stafford-Fraser
Interesting – but he article you reference is a rather poor skim of the original research.
The opening paragragraph mentions only Americans, which is common in American articles, aimed at Americans, but also pervades Wikipedia. It then states that sign language is only used to communicate between deaf people, which is clearly nonsense – a moment’s thought tells you it is used by people with hearing or speaking difficulties to communicate with everybody else.
The article then concentrates on the glove, perhaps the simplest, but most easy to grasp (!) part of the project. Actually, it is the machine learning system, underlying the translation, which is new.
Well, new-ish. A quick search reveals many such systems, mostly aimed at understanding sign language from video, rather than requiring the signer to do anything special – and not just American Sign Language!
Yes, the article’s poor, but the idea’s good, I think. Much easier to do this with gloves and get a high degree of accuracy than with video, for the time being, at least.
And I think they can be forgiven for focussing on Americans: they’re an American University and different sign languages are used elsewhere; so they naturally picked ASL as a starting point, and I’m guessing that ASL has one of the biggest user populations. If I were doing speech recognition, I would start with English and so it would be natural for any article to talk about English-speakers.
I did notice, though, the slight confusion about whether sign language is primarily used by people who are hearing- or speech-impaired! Still, they are correct about it only really being used in that community, because they and their helpers are the ones who understand it, so a translator for the rest of us is useful.
I immediately went to look at the actual paper itself, but it is, sadly, behind a paywall.
Well, at least they included one sentence about how translation gloves are generally viewed by the Deaf as misguided efforts by engineers who treat Deafness as a disability. Sadly, that’s better than a lot of media coverage. I definitely recommend clicking through to the Atlantic article they link.
Ah, thanks, Spencer – I’d missed that.