Urban75 Home About Offline BrixtonBuzz Contact

Smart glasses allow deaf people to ‘see’ conversations with subtitles

Petcha

Well-Known Member
This is pretty amazing. I saw an interview with the inventor earlier who was watching his Dad watching TV with subtitles and thought why the hell can't we just put combine voice recognition with AR/wearable tech and produce subtitles in real time while people are chatting. I'm sure many people have used translate features on their phones now and then while travelling to ask simple questions so this is an obvious progression.

A brilliantly simple idea and one which I'm surprised nobody came up with earlier - did Google Glass do this? The reaction from the deaf community has been amazing apparently. That video's quite good.

 
Indeed. They also look far cooler than the Google Glass shit with that stupid camera on the front. I assume they're working on something to recognise the speech patterns of those you often talk to, family members etc.
 
This is one of those ideas that will be great when they iron all the kinks out of it, In the meantime it is going to have to contend with dropped network connections and poor reception. Plus of course you've only got to watch the subtitles on the BBC News to realise that automated real time subtitling has some way to go yet.
 
Yep speech recognition is definitely getting better. It’s coming to apps like zoom and teams. Might already be in them actually I haven’t checked.
 
not sure if that's true - presumably the voice to text thing can be done using a smartphone locally rather than via the internet?

Not yet. This stuff still relies on cloud services. Although there are apps like DragonDictate but you need to train them to your voice.
 
Speech recognition is scarily good these days. I stick it on youtube videos sometimes and it's amazing how accurate it is even with noise/accents.
Yes. I remember getting a computer in the mid-late-1990s that purported to have voice recognition, so you could use simple commands to open and save files and navigate menus and so on. It wasn't very good though, would only understand when I tried using the calculator function if I said something like 'two plus two equals' in a fake American accent. My English accent isn't a strong regional one. When I'm down south, people can tell I'm a northerner because of some flat vowels when saying bus or cup or mug or bath or whatever, but here up north some people wonder if I'm a southerner. So it's not like I had a strong accent that you might expect to have baffled the software.

A few years later, I participated in a trial at work of transcription software, which again needed lots of tidying up and corrections.

---

But now voice recognition is so good that I can even dictate this post. I'm using the dictation feature on my mobile phone to write this. It's very good so far hasn't made any mistakes as you can see. He even allows me to punctuate. When I say 4 stop it types a. . Oops there is a glitch in the Matrix.
 
Yes. I remember getting a computer in the mid-late-1990s that purported to have voice recognition, so you could use simple commands to open and save files and navigate menus and so on. It wasn't very good though, would only understand when I tried using the calculator function if I said something like 'two plus two equals' in a fake American accent. My English accent isn't a strong regional one. When I'm down south, people can tell I'm a northerner because of some flat vowels when saying bus or cup or mug or bath or whatever, but here up north some people wonder if I'm a southerner. So it's not like I had a strong accent that you might expect to have baffled the software.

A few years later, I participated in a trial at work of transcription software, which again needed lots of tidying up and corrections.

---

But now voice recognition is so good that I can even dictate this post. I'm using the dictation feature on my mobile phone to write this. It's very good so far hasn't made any mistakes as you can see. He even allows me to punctuate. When I say 4 stop it types a. . Oops there is a glitch in the Matrix.

How do you get it to do the punctuation? It's been ages since I tried it, but seem to remember this was the main problem.
 
We use the auto transcription on Zoom a lot for live subtitling on virtual/hybrid events at my work.

It’s pretty amazing - even often going back and self editing previous words based on the context of later ones. Where it falls over is with any sort of technical/niche stuff, or certain strong regional accents, which is to be expected I guess.

The rate the technology has progressed is highly impressive and it’s only going to get better.
 
I use otter.ai recommended by someone on here as I recall. It seems really good although I've not had to go through and check yet. And yes highly impressive, hadn't heard about the Zoom auto transcription.
 
Teams also has captions now. In fact so does PowerPoint, the PowerPoint ones are very good. None of them are as good as human captions. And you can have your phone calls captioned, haven't tried that.

These stories slightly annoy me. Newspapers /news sites love stories about some technological widget that's going to "cure" disability and the reality of disabled people's lives is usually a lot more complex. Often the thing is a minor advance with some drawbacks. Like cochlear implants for instance.

In fact there is a word for this in disability circles, a "disability dongle".
 
Back
Top Bottom