The concept behind the the new device is with following the way the brain works when someone is engaging in a conversation in a crowded space, such as at a train station or at a party. The brain has a natural ability to single out and amplify one voice against a range of other human voices, which form background conversation, and from an array of other sounds. A design-limitation with hearing aids, even those capable of boosting sound and to the clearest levels, is that they function by boosting all voices equally. This works well in quite spaces but serves the hearing-impaired wearer badly when worn in crowded environments.
This problem has been overcome with the development of a new type of hearing aid from the Zuckerman Institute at Columbia University in New York. The new device utilizes artificial intelligence together with sensors to screen out unwanted noise. The device achieves this by monitoring the listener’s brain activity.
In doing so, the device adapts to each user. A specially developed algorithm is trained to separate the voices of multiple speakers. The artificial intelligence then compares these voices as audio tracks and compares them to the brain activity of the listener. The AI is able to zero in on which person the user is paying attention to. This is because the device wearer’s brain activity tracks the sound waves of one voice more closely than others. From this analysis, the speaker whose voice pattern most closely matches the device user’s brain waves becomes amplified over all other voices and sounds. This enables the user to tune into the person that they actually want to hear.
The video below provides more detail about the device:
Commenting on the new development lead researcher Nima Mesgarani told The Guardian: “The brain area that processes sound is extraordinarily sensitive and powerful. It can amplify one voice over others, seemingly effortlessly, while today’s hearing aids still pale in comparison.”
The development is described in the journal Science Advances, with the paper headed “Speaker-independent auditory attention decoding without access to clean speech sources.”