Noise-canceling headphones excel at drowning out background noise, but selectively allowing desired sounds remains a challenge. While existing technologies like Apple’s AirPods Pro attempt to adjust sound levels automatically, they lack user control. However, a breakthrough AI system developed by a University of Washington team offers a solution.
#### Target Speech Hearing System
Presented at the ACM CHI Conference on Human Factors in Computing Systems, the “Target Speech Hearing” (TSH) system enables wearers to focus on a specific speaker in a noisy environment. By simply looking at the speaker for three to five seconds, the system “enrolls” them, cancelling out other sounds and playing only their voice in real-time through headphones.
#### How It Works
Users wearing standard headphones equipped with microphones initiate the process by tapping a button while facing the speaker. The system, powered by on-board machine learning software, learns the speaker’s vocal patterns from the captured audio. Even as the wearer moves around, the system maintains focus on the enrolled speaker, continuously refining its accuracy with additional input.
#### Performance and Future Prospects
Tested on 21 subjects, the TSH system received high ratings for clarity compared to unfiltered audio. While currently limited to one enrolled speaker at a time, the team aims to expand its capabilities to earbuds and hearing aids. The code for the proof-of-concept device is open-source, encouraging further development and innovation in the field.
This groundbreaking technology offers a glimpse into the future of personalized audio experiences, enhancing communication and accessibility in noisy environments.
