Artificial intelligence could be the key to helping hearing-impaired people tune out the din and hear what's important, according to Google and Australian researchers.
The tech giant this week revealed work on its Australian Future Hearing Initiative had progressed to training hearing aids and cochlear implants using AI technology and could be ready for trials soon.
If successful, the technology could improve results from both devices and help users navigate noisy environments.
The hearing research is one of several AI health projects being undertaken by Google, with others focusing on detecting diseases such as tuberculosis and breast cancer.
Google senior research scientist Simon Carlile said he began investigating the impact of advanced technologies hearing devices after sitting beside his father at a family gathering.
"When I asked him why he wasn't wearing his hearing aids, he said that in noisy situations like this they just don't help," he said.
"It made me very sad at the time that where he really wanted to and where he really needed to interact with people, with his whole family around him, he couldn't."
Hearing aids could make noisy environments like parties and restaurants difficult to navigate for users, Dr Carlile said, as they amplified the noise around them without isolating important, nearby conversations.
To investigate whether technology like AI could assist, Google teamed with Australian research groups including Cochlear, Macquarie University, NextSense and The Shepherd Centre in two projects.
Cochlear director Zachary Smith said the first project used computational modelling and deep learning to compare normal hearing to a user's impaired hearing, and train a hearing aid to adapt to the difference.
The project, he said, had advanced to early testing.
"Right now, we have (the) first algorithms successfully training for both hearing aids and cochlear implants and we hope to start listening experiments in the near future," he said.
"Our hope is that such a machine-learning algorithm will sound more natural."
In the group's second hearing project, AI tools are being used to identify and artificially remove unwanted sounds to help hearing-aid users follow conversations in challenging environments.
"Complex listening environments that exist in everyday situations may include dozens or even hundreds of signals from multiple sound sources that overlap," Dr Smith said.
"We are quite excited from a scientific and researcher's perspective about the possibilities of artificial intelligence and machine learning."
Dr Carlile said the researchers would continue to improve the algorithms before testing the technology more widely, but he hoped the projects would "move the needle" on hearing-aid technology.
Google is also using AI technology to power research in other health projects, with the company revealing it had trained a model using 300 million coughing audio clips in a bid to detect early signs of disease, and was using AI models to detect breast cancer on ultrasounds in a Taiwanese hospital.
