Bringing a hearing aid into the factory: does $200 AI noise reduction hold up?

Depends. For daily use by an elderly person with moderate-to-severe hearing loss, I think it's worth it at this price; but if you're going in expecting "put it on and hear like a normal person," you'll be disappointed.

My old man has had hearing loss for years. We bought a few-hundred-yuan "sound amplifier" before — essentially it just amplifies everything together, so watching TV sounds like an argument, and going out he still can't hear clearly. This time we got him the Feisheng H9 Nature Pro to try for over half a month, and I also took it into the workshop for two afternoons.

It's a behind-the-ear model, BTE in industry terms — the main unit hangs behind the ear, with a thin tube running to an earpiece inserted in the ear canal. This type has more power than the small devices that sit inside the canal; moderate-to-severe hearing loss is its main battlefield.

Fitting is all done in the phone app. First a self-administered hearing test: put the earpieces in, tap the screen when you hear a beep. The first time we did it in the living room, the two results differed a lot, and the old man got annoyed. Later we moved to the bedroom with the door closed, and it stabilized once it was quiet. We got stuck on this step for almost forty minutes — dirty environment was the main cause. I've said similar things about QC data: if the input is dirty, whatever you compute afterward is wrong.

Inside is a 12nm AI chip, claiming 5.4ms latency. I'm familiar with latency — vision cameras look at this metric too; high latency gives you an echo sensation when you speak, very awkward. He didn't mention it in the first few days of wearing it, so I judge this item passes.

The most obvious change at home is the TV volume. It used to be turned up loud enough for the neighbors to hear; now at a normal level he can follow along. When multiple people talk, voices really do get pulled out, and background range-hood and fan noise gets pushed down — the advertised "noise reduction without reducing voices" largely holds up in my testing.

The pitfall is feedback squeal. If the earpiece isn't seated snugly, sound leaks, the device picks it back up and amplifies it, producing that sharp whistle. Swapping earpieces has to be tried repeatedly — the stock ones aren't necessarily right, don't expect to get it in one go; we only settled down at the third pair.

About the workshop, I should be clear: I had him stand next to the assembly line for about twenty minutes. It handles steady machine roar, but sudden metal clanking still makes him jump. A hearing aid is not hearing protection — for industrial noise you still wear earplugs, don't mix the two up. That's my own judgment, and it's also why I brought it into the workshop — I just wanted to see if anyone treats it as protective equipment.

On the plus side: pricing starting at 1999, a much lower barrier than those imported brands; it's domestic, so fitting doesn't require a trip to a store; voice enhancement is genuinely worked on. On the downside: self-administered hearing test accuracy isn't as good as an audiologist's equipment; earpieces take several rounds of trial; some settings in the app are buried in submenus — took me ages to find them. People with mild hearing loss will find it too loud; for complex cases like single-sided loss or profound loss, I wouldn't dare draw conclusions for them — still need a professional fitting.

The material mentions that in China only about 5% of people with hearing loss wear aids, blocked by two things: price and fitting hassle. This wave — Signia, Widex, Phonak, Starkey — are all pushing AI, same direction. My next step is to have the old man wear it for a full month straight, to see long-term comfort and whether the battery holds up. Short-term usability and long-term willingness to wear are two different things.

It's a good thing, but it can't replace the audiologist, and it can't replace the earpieces.