Hacker News Reader

Comment by localhost | Hacker News Reader

My dad explained to me that his hearing loss is dominated by his inability to zero in on conversations in a crowded room. That has made it so that he’s much less willing to go out to socialize in restaurants, because that environment creates a form of isolation brought on by his hearing loss. While his hearing aids certainly do help, I would argue that they haven’t gone digital enough.

Being able to audibly isolate a speaker in a room full of other speakers would be a boon for his case (and I would presume many others). Anyone on HN working in this area?

View on HN

Speech enhancement (denoising/dereverb) and seperation is absolutely a hot research topic, as it's needed for fields like voice assistants as well. Unsurprisingly, the state of the art is variety of deep neural networks, here's a good overview of recent developments [1]. Of course, many of these are far from real time, but earlier techniques are used in some digital hearing aids iirc. Apple have also released [2] some details of the system the Homepod uses, along with audio samples.

A quite interesting further development of these is this paper [3] which uses brain activity to determine which particular speaker the listener is trying to hear, and then optimises that.

[1] https://arxiv.org/abs/1708.07524 [2] https://machinelearning.apple.com/2018/12/03/optimizing-siri... [3] https://advances.sciencemag.org/content/5/5/eaav6134

Deep neural networks should be able to do source separation.

Imperfectly, but the results are at least passable.

A pair of Sony WH1000 does a pretty good job with it's "voice mode" (which I assume is just some smart frequency filtering). I don't see why it would be that hard to make the same tech available in a smaller package.

Bose has a “hearables” line in this vein. I know 5 or 6 folks in their labs, and it’s sounding (hah) promising. They aren’t medical devices, yet, but beamforming earbuds are a thing now.

I expect to see full medical devices that do this in the next decade or so.

DeLiang Wang's group at Ohio State is doing great stuff for speech extraction in complex environments, as are some of my colleagues in other universities in collaboration with Starkey.

It’s coming along!

There is a lot of cool research in separating sound sources in general. From friends in the industry, I know that the traditional approach taking advantage of phase difference between microphones is hard, because the of the small separation distance (within 1 hearing aid device and even when you have one in each ear). Conversely I haven't fully comprehended it yet, because I've similarly read that Google Android's beamforming microphone tech should be quite good: https://github.com/shichaog/WebRTC-audio-processing/tree/mas...

Here are some examples of new research with neural networks https://www.youtube.com/watch?v=FMEk8cHF-OA og https://www.youtube.com/watch?v=zL6ltnSKf9k

Most restaurants also seem to be much louder now than they were 20 years ago. They're so loud even people with good hearing have trouble having a conversation. I worry that just going to a popular restaurant nowadays is damaging my hearing, and I'm starting to avoid the loudest ones.

As costs go up and fashions change as does the interiors of restaurants. Gone are the "soft" things like plants, carpets, dropped ceilings, and plush furniture. More common these days is hardwood, cement, steel, glass, and exposed ceilings.

It's no coincidence that louder often means spending more on drinks and quicker table turn overs. So the noisy places are out competing the quieter places.

Sad, I find about half and 75% of places are unacceptably loud during normal peak hours. Fortunately google displays how popular a place is (if often varies by business) so I can go off peak.

My Airpods have a mode where you can use them for directional hearing. You can set the iphone next to the person speaking and whatever it picks up is transmitted to the airpods. It would be nice for hearing aids to be able to wirelessly connect to a directional mic for situations like this.

His hearing aids can already connect over BT. He uses it a lot to watch TV. But I just checked and there seems to be a bunch of apps on the App Store that do some variation on this idea. Something to try out with him on my next visit home. Thx for the idea!

This is sometimes called the “cocktail party problem”, or more formally “Blind Source Separation”. It’s tough, especially with one or two sensors. You’d also need a fairly unobtrusive way to choose the source, which is also tricky.

That said, I had a classmate with a hearing impairment. His device had a tiny wireless microphone that preferentially fed his hearing aid (I.e., he could hear other stuff too, but the mic was loudest). The teacher usually wore it, but it would be given to you if you were working with him on something. It was certainly a little clunky, but it did seem to help him a lot. Perhaps something similar could work for your dad?

Unfortunately, I have no idea what the device was called—and it was a long time ago- but maybe this is enough of clue for his audiologist.

That’s a cool idea. I wonder if a phone could be that device in today’s world - something you would bring out and drop onto the table to act as another set of microphones and perhaps more importantly, a lot more compute power?

I’ll pass that tip along to him. Thx!

When was the last time your dad tried a newer hearing aid? Some now have multi-directional microphones that are meant to solve for the use case you described. They've been getting some good reviews, and with the recently de-regulation expect the pace of change/competition to get more fierce.

Pretty sure he bought his last one at Costco ... a year ago? It’s frustrating because adjustments require trips to his audiologist and he often complains that it doesn’t seem to improve things much. It would be great to have one that enabled self-tuning, but can also understand why that could be a terrible thing for some people.

Self fitting is definitely an option - check out this link: https://forum.hearingtracker.com/c/hearing-aid-self-fitting-...

Alternatively, let me know and I can walk you through the process.

I recall hearing (so take this with a grain of salt), that part of the issue is when you start looking your hearing you stop exercising the skill of focusing on a single stream of sound, such that when you get your hearing back you ability to do it has greatly diminished.

Yes, directionality is an issue. But the brain post processing portion has also dwindled.

So i guess getting a hearing aid sooner might be a good approach?

Is that a problem of focus or hearing?

“Yes”

It’s actually hard to separate the two. On the one hand, the hearing aid might not faithfully reproduce cues that let you separate different sources. On the other, “higher” brain areas can modulate early sensory areas (maybe even the cochlea) to make sounds relevant to your current behaviors more salient while attenuating irrelevant signals.

As far as I understand it, the latter provides information needed for the former. It's possible to have deficits in either or both:

1) Hearing loss in specific volume/frequency notches instead of loss strictly at the high/low end (sometimes referred to as "hidden hearing loss", which is more a statement about the capabilities of routine screenings than it is a statement about the nature of the hearing loss)

2) Deficits in attention regulation, which are seen in numerous conditions (most obviously ADHD, but also autism spectrum, PTSD, traumatic brain injury, schizophrenia and "schizophrenia-like" conditions, and some forms of dementia)

Reply on news.ycombinator.com