Busan National Univ. Team Develops Noise-Cancelling Tech with Meta for AI Glasses

At a glance

A research team led by Professor Lee Dong-heon at Pukyong National University developed Virtual-Flow, a noise-cancelling technology for small devices such as AI glasses, in joint work with Meta. It delivers voice quality close to that of a 16-microphone setup using only four microphones.

Key points

1. 1

   It reduces surrounding noise and echo in small devices with few microphones.

2. 2

   It creates virtual microphone information without collecting large amounts of labeled data.

3. 3

   It runs in real time on devices by making computation efficient in a single pass.

4. 4

   It performed well on real recordings and in settings where the speaker moves.

Why it matters

It could improve call and voice recognition quality in small devices such as AR glasses and hearing aids.

Also reported by

• 아시아경제국립부경대 이동헌 교수·Meta, AI 글래스용 노이즈 캔슬링 기술 개발
• LG헬로비전국립부경대 이동헌 교수, Meta와 ‘AI 글래스용 노이즈 캔슬링 기술’ 개발