They’re effective at quelling the sounds of crying babies, but noise-cancelling headphones can also tune out sounds you might actually want to hear, like the horn of an approaching car as you absentmindedly step out into a busy street. To make noise-cancelling headphones safer, researchers are now teaching them to recognise and alert the wearer to sounds that represent imminent danger.
There’s no denying we now live in a world that’s very good at distracting us with technology, and pairing smartphones with headphones that are now remarkably effective at tuning out the sounds of the world is a genuine cause for concern. There have been some approaches taken to make noise-cancelling headphones smarter and safer, such as letting especially loud sounds through with the assumption they’re worthy of note, but it’s missing the quieter sounds we don’t realise our ears pick up on that are a bigger concern: like the sounds of the engine of an approaching vehicle.
Researchers from the Data Science Institute at Columbia University have come up with a possible solution to this first world problem. They’ve upgraded a standard pair of over-the-ear noise-cancelling headphones with additional microphones pointing outwards from the sides and back of the wearer’s head. These microphones work independently of the mics responsible for detecting and cancelling out noise, and the sounds they capture are routed to an additional app on a connected smartphone for processing.
Using machine learning, the app will be trained to not only recognise and differentiate the cacophony of sounds being picked up by the headphones but will also be able to assess the potential threat of each one. The sounds of a street busker playing music, or the beeps of a crossing signal for the visually impaired wouldn’t pose a threat, but the sounds of a motor getting louder and louder, implying it’s approaching, are an obvious concern which would trigger audible alerts sent to the headphones.
The researchers are currently testing their headphones on the pedestrian-unfriendly streets of New York, but are not only concerned with how well they detect dangers. The team is also working with a psychology professor to determine what audible warnings would be most effective at grabbing a user’s attention, alerting them with enough time to avoid danger, while not becoming something that would eventually be tuned out, eliminating the effectiveness of this project.
Featured image: Columbia University Data Science Institute