A University of Washington team has developed an artificial intelligence system that allows a person wearing noise-canceling headphones to enroll a speaker by looking at them for a few seconds. The system cancels out all other noise in the environment and plays only the enrolled speaker’s voice in real time, even as the listener moves around and no longer faces the speaker. This system, called “Target Speech Hearing,” builds on the team’s previous work with “semantic hearing” for noise-canceling headphones.
The AI system relies on off-the-shelf headphones fitted with microphones. By tapping a button while facing the speaker, the person wearing the headphones allows the system to learn and recognize the speaker’s vocal patterns. The system then continues to play back the enrolled voice to the listener in real time, even as they move around. The system improves its ability to focus on the enrolled voice as the speaker continues to talk, providing more training data for the machine learning software.
While there are noise-canceling headphones available on the market that can adjust sound levels during a conversation, the UW’s prototype takes it further by allowing the user to control whom they are listening to. By enrolling a specific speaker, the user can hear only that person’s voice, even in noisy environments like crowded restaurants or cafeterias. The system can currently only enroll one speaker at a time and requires there to be no other loud voices coming from the same direction as the target speaker’s voice.
The team presented their findings at the ACM CHI Conference on Human Factors in Computing Systems in Honolulu and made the code for the proof-of-concept device available for others to build on. The system is not yet commercially available, but is an exciting development in the field of artificial intelligence and noise-canceling technology. With further advancements and improvements, this technology could have a significant impact on how people hear and communicate in noisy environments.
