Noise-cancelling headphones are great for blocking out the loud talkers and chaotic coffee shops of the world, but what if there are certain sounds that you want to hear? That's where the experimental new "semantic hearing" system could soon come in.
Currently being developed by scientists at the University of Washington, the technology incorporates deep-learning algorithms running on a smartphone that's wirelessly connected to a set of third-party noise-canceling headphones.
Ambient audio is streamed from the headphones' noise-monitoring mics to an app on the phone, where it's digitally filtered to block most sounds but allow a select few to be heard by the listener. Conversely, it can also block select unwanted sounds while letting all others through. There are currently 20 such allowable and/or blockable sound choices.
For now at least, the smartphone definitely has to be part of the package. That's because the electronics in the headphones themselves – or on a cloud-based server – wouldn't be powerful or fast enough.
"Understanding what a bird sounds like and extracting it from all other sounds in an environment requires real-time intelligence that today’s noise-canceling headphones haven’t achieved," said Prof. Shyam Gollakota, senior author of a paper on the study. "The challenge is that the sounds headphone wearers hear need to sync with their visual senses. You can’t be hearing someone’s voice two seconds after they talk to you. This means the neural algorithms must process sounds in under a hundredth of a second."
When tested on 22 volunteers in settings such as parks, streets and offices, the semantic hearing system successfully extracted target sounds like bird chirps, alarms and sirens. Some work is still necessary, however, as the technology has difficulty differentiating between similar sounds such as human speech and singing.
You can hear the effects of the semantic hearing system for yourself, in the video below.
And if you just can't wait for the technology to be commercialized, you might want to check out an existing device known as the SoundBrake. While it doesn't filter specific types of sounds, it does filter sounds based on their volume.
Source: University of Washington
