Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Deep Learning enables hearing aid wearers to pick out a voice in a crowded room

spectrum.ieee.org

Post reply on HN
Print view View original

21–30 of 60 posts

Re: Deep Learning enables hearing aid wearers to pick out a voice in a crowded room

#21

Post by sdrothrock »

Kluny wrote: post #5
sdrothrock wrote: post #4

Earlier quoted context omitted.

As someone with a cochlear implant who lives with the consequences of overly clever programmers who thought they'd "help" by filtering out noise and volume and whatever else... I really wish they wouldn't. This is a technology that makes me so angry some days that I sometimes wonder if it was worth getting implanted, even though I know it was.

Which one have you got? I know what you mean with the "helpful" bullshit. My conventional hearing aid, on my left ear, has this "smart" mode where it tries to detect speech vs noise, and change the volume or the directionality of the microphone to compensate. You end up with this wildly fluctuating volume all the time where it feels like stationary objects are coming at you. I had them turn that feature off asap. On…

I have the Advanced Bionics Harmony BTE. Since my implant is AB, I wouldn't be able to get the Nucleus Freedom 6.

I have an in-ear mic, which does wonders for reducing surrounding noises and also for letting me use a phone normally, but my main issue is with the software itself; I've had issues with it since implantation and they've always been pooh-poohed by audiologists at Hopkins, Tokyo University, and Toranomon. The biggest problem is that it seems to operate on some kind of averaging system -- when there's a noise that's louder than the recent average, everything just cuts out for a few seconds. This is especially noticeable in the morning, where I've just woken up and am trying to get to work, but there are cars and trains etc. making noise and making my hearing cut in and out constantly, which not only drives me up the wall but gives me a terrible headache.

Re: Deep Learning enables hearing aid wearers to pick out a voice in a crowded room

#23

Post by KineticLensman »

chx wrote: post #9
tobtoh wrote: post #6

> The greatest frustration among potential users is that a hearing aid cannot distinguish between, for example, a voice and the sound of a passing car if those sounds occur at the same time. The device cranks up the volume on both, creating an incoherent din. It may be a simplification of the article that I'm misinterpreting, but as someone who got a hearing aid in early 2016, that's not how (modern) hearing aids wor…

I probably need a hearing aid but I am very reluctant to go for a testing because the last time I had one (very long ago and very far away from Canada, I admit) I thought it was incredibly imprecise: they asked me to press a button when I could hear a sound and I was absolutely unsure whether I heard really something already or just imagined. Is it still so very subjective?

Moderately deaf Brit here. The NHS used this technique on me a year ago when I got my hearing aid. Whether it is imprecise or not I don't know, but it generated a graph plot showing, for each ear, a plot of my hearing response against a range of different frequencies, which was annotated with the position of specific phonemes - so I could see which vowels, etc were hardest to hear. This seemed to match my experience, although I didn't test it rigorously.

If you are reluctant about getting a hearing aid, I would totally totally recommend it. It turned my social and professional life around - I was beginning to avoid conversation with certain softly-spoken people and couldn't follow arguments in conference rooms that had any sort of noisy aircon. As a Brit, I have access to free NHS hearing aids and consumables (batteries and the tiny tubes that go into the ear), which helps. I had been planning to buy a smaller in-ear device, but in fact the external device is acceptable in terms of size and appearance (silvery grey).

The device I use has an external control button (a tiny stud) which controls volume but which could alternatively be programmed to trigger different modes (e.g. noisy room vs quiet room). The fact that I almost never have to change the settings also suggests that the initial test (which is programmed into the hearing aid) was somewhat accurate.

Re: Deep Learning enables hearing aid wearers to pick out a voice in a crowded room

#24

Post by Eridrus »

chas wrote: post #7

This approach surprised me. Why are they doing feature extraction and then feeding that into a DNN? It seems much more straightforward to have the input of the network be noisy samples and the output be clean samples a la super resolution[0] in images. They probably wouldn't want to use fully-connected layers in that instance, but I don't see any fundamental barriers if they have enough computational power to run a n…

The only thing I've seen run on raw audio are WaveNet models, and those are way too expensive to get into an embedded chip, no public real-time implementations exist, though Baidu had a paper which claimed real-time execution speed on some server class Intel CPU last week. They do mention that their CPU implementation could be parallelized too.

Re: Deep Learning enables hearing aid wearers to pick out a voice in a crowded room

#25

Post by andrewaylett »

chx wrote: post #9
tobtoh wrote: post #6

> The greatest frustration among potential users is that a hearing aid cannot distinguish between, for example, a voice and the sound of a passing car if those sounds occur at the same time. The device cranks up the volume on both, creating an incoherent din. It may be a simplification of the article that I'm misinterpreting, but as someone who got a hearing aid in early 2016, that's not how (modern) hearing aids wor…

I probably need a hearing aid but I am very reluctant to go for a testing because the last time I had one (very long ago and very far away from Canada, I admit) I thought it was incredibly imprecise: they asked me to press a button when I could hear a sound and I was absolutely unsure whether I heard really something already or just imagined. Is it still so very subjective?

You might be interested in https://hearingtest.online/ -- it's obviously not medical quality, but with a decent pair of headphones should help you to see what sort of issues you might have.

I'll re-iterate the point other commenters have made, though: even if you're not entirely sure you heard a sound, press the button. It's all useful data: if you're pushing the button during a high-frequency part of the test when there's actually no sound, that's a sign things are not entirely right. And if you only push the button when there is a sound, but don't push it half the time the sound is played, that's also useful data.

Re: Deep Learning enables hearing aid wearers to pick out a voice in a crowded room

#26

Post by radarsat1 »

chx wrote: post #9
tobtoh wrote: post #6

> The greatest frustration among potential users is that a hearing aid cannot distinguish between, for example, a voice and the sound of a passing car if those sounds occur at the same time. The device cranks up the volume on both, creating an incoherent din. It may be a simplification of the article that I'm misinterpreting, but as someone who got a hearing aid in early 2016, that's not how (modern) hearing aids wor…

I probably need a hearing aid but I am very reluctant to go for a testing because the last time I had one (very long ago and very far away from Canada, I admit) I thought it was incredibly imprecise: they asked me to press a button when I could hear a sound and I was absolutely unsure whether I heard really something already or just imagined. Is it still so very subjective?

Ah, I see.. interesting interpretation. It's a shame they didn't explain the process to you better.

This is how psychometric testing works. It's inherently difficult because in order to estimate the point of subjective "loss", which we call the "just noticeable difference" (or JND), one has to sample more in the area of the variable (amplitude, frequency, etc) that is more difficult for you to distinguish. Consequently, one will always walk away from such an experiment with an impression of having "guessed" and being really not sure if you gave the right answers. But that's because they're trying to estimate exactly that: they're trying to find the point at which you really aren't sure whether you hear something or not.

Basically this: if you guessed perfectly every time that you heard something, that would be a 100% recognition rate. If you always said with 100% certainty that you didn't hear anything, that would be a 0% recognition rate. So logically, the point of hearing loss occurs somewhere between those two extremes.

In order to determine more precisely where, the procedure has to "zoom in" on the point at which you answer correctly 50% of the time, a bit like a binary search. (Or sometimes they want 75% or the time, etc.) In any case, to do so, they need to sample the probability of you answering correctly or incorrectly in that region. This sketches out a probability curve, and then they can fit that curve and figure out the 50% or 75% point on the curve.

They'll sample using either a constant spacing method, random sampling, or a staircase method that adaptively moves towards the 50% point. The latter is more efficient, in the sense that it requires fewer answers from you, so that is what is often used in practice. However, by its nature it is also much more frustrating for the patient, because it will be sampling much more frequently in the region where you are "not sure" of the answer.

I'm really sorry they didn't explain this stuff to you, and allowed you to walk away thinking it was a badly done experiment!

Re: Deep Learning enables hearing aid wearers to pick out a voice in a crowded room

#27

Post by tobtoh »

radarsat1 wrote: post #26
chx wrote: post #9

Earlier quoted context omitted.

I probably need a hearing aid but I am very reluctant to go for a testing because the last time I had one (very long ago and very far away from Canada, I admit) I thought it was incredibly imprecise: they asked me to press a button when I could hear a sound and I was absolutely unsure whether I heard really something already or just imagined. Is it still so very subjective?

Ah, I see.. interesting interpretation. It's a shame they didn't explain the process to you better. This is how psychometric testing works. It's inherently difficult because in order to estimate the point of subjective "loss", which we call the "just noticeable difference" (or JND), one has to sample more in the area of the variable (amplitude, frequency, etc) that is more difficult for you to distinguish. Consequent…

Thanks for that fantastic explanation! It wasn't explained to me either when I had my hearing tests - i just assumed they were doing multiple tests to get an 'average' answer.
Post reply on HN