A small group of computer science and engineering students at the University of Washington have designed a new set of wireless headphones that use AI-powered technology to provide a better pattern of background noise cancellation during phone calls.
The headphones – which students dubbed ClearBuds, a great name in our opinion – work by monitoring two separate audio streams from the microphones on the left and right buttons to create a spatial audio soundscape, sending the data to a phone that runs a pair of neural networks to clean up background sounds.
The first neural network suppresses any sound other than the caller’s voice, while the second amplifies and improves the clarity of that voice. The result is a significant improvement in background noise removal, which students found 6.94 dB better than the Apple AirPods Pro.
Both neural networks were trained over the course of a day, utilizing the AI deep learning capabilities of Nvidia’s Titan desktop GPUs. Nvidia has made a point of emphasizing the potential of AI to improve games via DLSS, but also how we use technology in all areas of life, recently setting up a new program for companies to start implementing AI technology at the enterprise level. AMD is even getting involveddespite previous reviews to the contrary.
Analysis: Amazing technology that will need an industry shift to become useful
At this point, most wireless headphones implement some level of active noise cancellation (ANC), which works but monitors outside sounds through a microphone and creates an inverted sound wave signature to cancel them out. Some are very good, but this new technology could lead to a jump in ANC performance.
There are a few issues that need to be addressed first – and they are not issues with ClearBuds (again, great name) per se. The first is that the neural network requires spatial audio to function correctly, which means the user must use both earbuds for full functionality – and we won’t be the only ones to blame for putting a single button on when out and about.
A more pertinent issue identified by UoW students in their post-project review is the fact that consumer headphones currently use ANC monitoring audio from a single headphone, despite the fact that most feature a microphone on both sides. buttons (because of users’ propensity to use just 1).
That means manufacturers would have to change the way their devices work to implement this AI-based ANC, and the technology could also result in decreased battery life for the headphones and connected phone, which fuels the process. real neural network. Personally, we hope these students can find a business partner – with such a good name, ClearBuds sounds like the next big thing.