Hacker News Reader

Nice PR piece :)

We need dramatic improvement in battery capacity to put machine learning in hearing aid, it's far away

The article suggests any ML is happening in the paired smartphone, though it doesn't give details. I doubt there's enough room or cooling in the hearing aid itself, let alone battery.

That can't work because of the lag of the phone to hearing aid communication

It wouldn't necessarily need to be realtime - if you're using the device to build/tweak a "fingerprint" of background noise or required corrections, you could do that analysis with samples and periodically push the resulting filter profiles back to the earpiece (or something similar) after the phone has done the number crunching to generate them.

The smartphone just switches modes and adjusts the volume. AFAICT these modern hearing aids are doing something like dynamic EQ, based on frequency level detection. The programs that dynamically adjust the EQ levels are trained on test material and then the audiologist or the user can switch between programs and adjust the levels based on the user's hearing response curve.

So what happens in real time on the device is fairly simple.

AFAICT - someone please correct me if they know otherwise.

Reply on news.ycombinator.com