The digital hearing aid revolution that started in the ‘80s is on the brink of its next evolution: machine learning. What will it mean for end users? For you? We’ve taken a look in our crystal ball and have some predictions.
But first, a quick look at the present: Modern digital signal processing has made noise reduction, beamforming, feedback cancellation and speech enhancement standard in hearing aids.
It has also made them so complex that precisely adjusting them to individual needs can be challenging! Yet that’s what hearing care professionals do every day. And you’re good at it.
More ease, accuracy and time for more clients
Machine learning has the potential to minimise that complexity and improve your job satisfaction by making the in-clinic fitting process faster, easier and more accurate. It can also increase end-user satisfaction by helping your clients hear better during their daily, real-life activities – since you simply can’t follow them around all the time.
“Machine learning will add yet another important level of ease and hearing accuracy,” says Oliver Townend, Widex Audiologist and Audiology Communications Manager and co-author of the article Real-life applications of machine learning in hearing aids in The Hearing Review
. “It will also free up more time for hearing care professionals to see new clients or boost their business in other ways.”
Townend says machine learning is a natural fit, in particular because of the challenges people with hearing loss face out in real life.
“Hearing in real life constantly changes, because real life itself changes from one moment to the next. And having to think about what hearing aid program to use in a given situation requires cognitive resources. That’s why we’ve already built so much automation into our hearing aids – so people can use their cognitive resources on the task of listening instead of figuring out which program they should use,” Townend explains.
Existing automation systems have challenges
However, he points out, automation is built on assumptions about what to amplify and not to amplify. And, even the best available automated system cannot know what the user intends to hear in any given scenario.
For example, if a client is at a social gathering, does he or she always want to hear the conversation? Or would they sometimes like to hear more of what that great pianist is playing in the background?
Or think about the fitting situation
. Your clients try to explain how they perceive the sound or hearing challenge – which is not always easy for them. Then you have to interpret what they’re describing. Sometimes it’s an (educated) guessing game!
Furthermore, there are thousands of different hearing aid parameters that can be adjusted. If you wanted to compare whether the client thinks “Sound A” is better than “Sound B” on a typical hearing aid, it would take nearly 2,500,000 comparisons to get through all the parameters!
Enter machine learning
In light of the above challenges, Widex suggests a hearing aid solution that is built on machine learning and is driven by the user’s preferences and intentions.
Townend explains: “We propose a simple interface that uses the hearing aid user’s smartphone. Step by step, we can guide the user to better hearing by using simple A/B comparisons. They just choose what sounds best each time. Then, machine learning helps us predict their preferred setting. All the client has to do is choose what sounds best, based on what it is they intend to hear. The machine learning algorithm takes care of the rest.”
Widex research has shown that such a system is capable of reaching an ideal hearing aid setting within approximately 20 comparisons, as opposed to nearly 2,500,000.Read more about machine learning in hearing aids here >