“Do you have the AI hearing aids?”
Louai and I are asked this most weeks now, and it deserves a better answer than simply yes or no.
AI has genuinely changed what some of the newest hearing aids can do. But it has also become a label attached to almost every new piece of technology, which can make it difficult to work out what you are actually paying for — and whether it will make any meaningful difference to your hearing.
So let’s take away some of the mystery.
What is AI actually doing in a hearing aid?
One of the things we hear most often in the clinic is:
“I can hear people talking. I just can’t always understand what they’re saying — especially when there are other people or noise around.”
Usually it is the restaurant. The family lunch. Friends around a table. The car. A meeting where people keep speaking from different directions.
And this is where some of the newer technology is becoming particularly interesting.
AI can now be used in hearing aids to recognise different listening environments, identify speech and different types of noise, decide how the microphones should focus, reduce unwanted background sound and automatically change the way the hearing aid processes sound as your surroundings change.
Some systems can follow several people taking part in a conversation. Others take information about your movement or the acoustic environment into account. Some use AI in their phone apps to help personalise sound according to your preferences.
All of those things can be useful.
But, for me, the most exciting change is still one of the simplest to explain:
Hearing aids are becoming much better at separating people’s voices from all the other sound happening around them.
And that matters because this has always been one of the hardest problems for hearing aids to solve.
Why is a noisy restaurant so difficult?
Think about sitting at a restaurant with six friends.
The person directly opposite you is talking. Someone beside you joins in. There is another table behind you. A waiter asks a question from your left. There is music playing, cutlery hitting plates, chairs moving and the coffee machine going in the background.
Your brain is remarkably good at deciding which of those sounds deserves your attention.
A hearing aid has a much harder job.
For many years, manufacturers dealt with this by giving hearing aids a set of clever but fairly rigid rules.
Sound in front of the wearer is probably important.
Steady sound like a fan, traffic or air-conditioning is probably noise.
Sound from behind might be less important, so reduce it.
Those rules still have value and are still used in modern hearing aids. Directional microphones in particular remain an important part of helping people hear in noise.
The problem is that real conversations do not follow neat rules.
Your daughter does not remain directly in front of you for the entire family lunch. Someone calls your name from the side. People interrupt one another. You turn your head. The conversation moves.
Real life is messy.
This is where AI changes things
With newer AI-driven systems, manufacturers can train a computer model using enormous collections of real-world sounds — speech, different voices, restaurants, traffic, household noise, music and countless combinations of speech mixed with noise.
The system is exposed to these examples again and again during development so that it becomes very good at recognising patterns that belong to speech and patterns that belong to other sounds.
A version of that trained system can then operate inside the hearing aid.
So rather than relying only on a rule such as “the voice in front is probably the one we want”, the hearing aid can analyse the sound itself and become much better at recognising:
“That is speech. That is speech too. That sound is background noise.”
Some of the newest systems can do this with voices arriving from several directions rather than only from directly in front of you.
The hearing aid can then reduce more of the competing noise while keeping the speech information available.
I often explain this as helping to lift the voices out of the background.
The restaurant does not suddenly become silent. Nor should it. You still want to hear the people, the atmosphere and enough of what is happening around you to feel part of the room.
But the aim is for the voices you are trying to follow to stand out more clearly from everything else.
Does the research show that it actually works?
There is now encouraging published evidence showing that deep-neural-network, or DNN-based, noise reduction can improve speech understanding, clarity and listening effort in difficult noise compared with more conventional processing in certain situations.
That does not mean every person gets the same result.
The benefit can vary depending on the type of noise, where the speaker is positioned, the degree and type of hearing loss and the particular hearing aid being used.
That last part is important.
You may see advertisements saying a hearing aid gives “X% better speech understanding”, “X dB improvement” or “X% less listening effort”.
Those numbers are not necessarily meaningless, but they are usually measured under very specific test conditions. They should not be interpreted as a promise that every person will experience exactly that improvement at their next family dinner.
I think the fairest way to describe the technology is this:
It is genuinely promising and can make a very real difference for the right person — but it does not perform miracles.
What AI does not do
This part is just as important.
It does not restore normal hearing
A hearing aid is still working with the hearing system you have.
If hearing loss has reduced access to certain sounds, or your auditory system has difficulty separating speech from competing information, technology can help — sometimes considerably — but it cannot simply turn your ears back into those of a 20-year-old.
It does not magically know who you want to hear
This is probably one of the biggest misconceptions.
The hearing aid can become very good at recognising speech amongst noise. Some systems can track several conversational voices around you.
But it does not read your mind.
If three people are speaking at once, it cannot always know that this is the person you have chosen to listen to.
You and your brain are still very much part of the process.
The main AI has not been personally “trained” on you
The deep neural network doing the speech-and-noise processing is generally trained by the manufacturer before the hearing aid reaches us.
That does not mean modern hearing aids cannot personalise anything. Some devices and apps can remember adjustments, learn listening preferences or suggest settings based on what you choose.
But that is different from the hearing aid continually retraining its underlying speech-separation system specifically on you.
It does not replace a good fitting
This one is particularly important to us.
You can put incredibly sophisticated processing into a hearing aid, but if the basic amplification has not been set correctly for your hearing loss, the technology is starting with the wrong information.
It is a little like buying an extremely clever pair of prescription glasses but putting the wrong prescription into the lenses.
That is why we use Real Ear Measurements to measure what the hearing aid is actually delivering inside your ear canal and adjust it against a prescription for your hearing.
More technology does not make good clinical care less important.
If anything, it makes it more important.
There is more to the AI story
There is actually so much to say about AI in hearing aids that it really deserves two articles — otherwise this one risks turning into a small novel!
In this first part, I wanted to explain what AI is actually doing inside a hearing aid, particularly why it is becoming so much better at helping separate conversations from the noise around them, and also what it cannot do.
But there is another important part of the conversation:
Not all “AI hearing aids” are doing the same thing.
Different manufacturers have approached the problem in different ways. Some concentrate heavily on separating speech from background noise. Others focus on following several people in a group conversation, keeping the surrounding environment sounding natural, adapting as you move, or combining AI with highly directional microphones.
So in Part 2, we will look at how the main hearing aid manufacturers are using AI differently, whether paying more for newer technology is worthwhile, how the Australian Government Hearing Services Program fits in, and how we decide which technology might actually suit you.
Because ultimately, the question is not:
“Which hearing aid has the best AI?”
It is:
“Which hearing aid is most likely to help me in the situations where I struggle?”
