Hacker News Reader

Comment by CodeCompost | Hacker News Reader

As somebody who is hearing impaired, a feature like this would be a Godsend for me! This feature should be integrated into hearing-aids ASAP! Shut up - no, actually - keep talking and take my money!

View on HN

And we almost all will be where you are now, if we live long enough.

If you can pick out audio from individuals, you could also send it through speech recognition and subtitle real world conversations for when hearing is worse or not there at all.

But it really needs mobile devices capable of doing the processing locally, as I think round trips to the cloud would make it less useful or potentially useless.

How much latency is acceptable? If you're off in the woods somewhere far from the cloud, sure, but it's less than 10ms to ping Google.com for me, and if the speech-to-text engine runs faster than realtime, I don't see why processing remotely is a problem. 10 ms is nothing.

Still, the transcription part is already here today. The Google Translate app has a transcribe app that does this (runs locally; does not do magic AI "pick voice out from crowd"). My father-in-law has been using it for years. When I'm in a loud environment, the app I use on iOS is called Big, which just displays large text on the screen.

https://apps.apple.com/us/app/make-it-big/id479282584

10ms on a wifi connection is exceptional; on a cellular connection it's unheard of. I normally get 70-80ms on 5G, which is well past the threshold for realtime—and that's with a solid connection.

> If you're off in the woods somewhere far from the cloud, sure, but it's less than 10ms to ping Google.com for me

I'm in one of the biggest cities of my current country, and the RTT to google from me is 87-91ms. Well over 4 million people live within 100km of me, so I suspect they see similar latencies. On my cell, I see 191-207ms.

That’s a shockingly high latency for a major city! Getting 3ms to google.com here, big city in the Netherlands. Probably close to a data center.

Also in a big city in the Netherlands, but I just blame ziggo. We're getting fiber in my neighborhood ... "soon" ... so we'll see how it is once that happens.

Looks like a good 60ms is nothing but buffer-bloat in the router, as when pinging directly from the router, the RTT is much less.

Latency or not, for privacy reasons.

> And we almost all will be where you are now, if we live long enough.

And given the US healthcare system, somebody is gonna take all our money too, one way or another. :P

"But it really needs mobile devices capable of doing the processing locally"

I would think this shouldn't be a problem as the correct hardware gets adopted in phones. As it stands now, you could probably run it on a Coral USB accelerator and battery run Pi (just an example of hardware, obviously we don't have the code).

>and subtitle real world conversations for when hearing is worse or not there at all.

Or for when you don't speak that language.

I have sensoneural hearing loss as well and fyi Bose Hearphones do have something a little like this with directional noise cancellation that helps a lot. They are discontinued but you can find them refurbished.

My phonak HAs have some directional noise cancellation (or biasing at least; I don't have rigorous definitions for these terms)... It helps but isn't great.

Has a problem that I think the AI headphones wouldn't solve either: in a (non-quiet) group setting you still need to anticipate who's going to speak when and look at them for best results.

The direction bit is just biasing to preferring forward stuff (via two mics on each ear's HA).

Sadly, no backwards bias option for overhearing people behind you ;)

Version 3 will be able to analyse a room for interesting conversation and then control where you are looking via neuralink.

> Sadly, no backwards bias option for overhearing people behind you ;)

Put them on "backwards"; left cup on right ear and vice versa: forward facing mics now face backwards ;)

It's a bit physically trickier than that due to curved tubing and ear molds... but I could totally "try" it with friends, e.g. rotating the BTE hearing aid 180 so it's forward, and they'd have fun too.

The Bose have 2 settings for this, 180 degrees frontal, and a much narrow directly in front of you.

My Sony's have a "focus on voice" setting in the noise cancelling section of their app. Is it similar?

I haven't tried those but sounds like possibly just adjusts frequencies vs using directional mics. Might be same as Airpods Pro which I should try.

You should look into Luxottica's efforts in this category. Wearable glasses are quite promising for the use case you mentioned, as they avoid the bulk and impoliteness of wearing headphones while talking to someone.

>> https://www.cnet.com/health/medical/what-did-you-say-these-e...

Reply on news.ycombinator.com