ClearBuds' real-time noise-suppression technology outclasses Apple AirPods Pro
A paper on ClearBuds wireless binaural earbuds has been shared at Mobisys 2022 in the US. ClearBuds uses a neural network to enhance the speech streamed from two wireless earbuds. It promises real-time sound separation and background cancellation, both in high quality. But how good is it? Well, the real-time noise suppression technology even defeats AirPods Pro, Apple's much acclaimed flagship earbuds with ANC.
Why does this story matter?
Wireless earbuds have become extremely popular worldwide as they make taking calls on the go convenient. However, background noises make it difficult for both the speaker and listener to communicate. ClearBuds takes care of this issue by suppressing background sounds using speech captured across two earbuds. This is a revolutionary technology and should rake in moolah when it heads to the market.
It has a synchronization error of less than 64 microseconds
ClearBuds makes two technical changes to bridge the gap between in-ear mobile systems and deep learning for blind audio source separation. Its design makes it able to operate as a binaural synchronized microphone array and its dual-channel speech enhancement neural network can run on mobile devices. Tests show that the earbuds have a synchronization error of fewer than 64 microseconds.
ClearBuds are perfect for use in a noisy environment
Understanding the technology
ClearBuds is designed to isolate a person's voice when other people are talking or when there is a background noise. This separation of voice and noise is done by channeling audio from both earbuds to your smartphone where a neural net removes background noise in real-time. When taking calls using ClearBuds, you just have to turn on noise suppression option on your phone.
ClearBuds' noise suppression is much better than Apple AirPods Pro
Who are involved with this project?
Seven people have collaborated on this wireless binaural earbuds project. They include Vivek Jayaram, Ishan Chatterjee, Maruchi Kim, Shyam Gollakota, Steve Seitz, Shwetak Patel, and Ira Kemelmacher-Shlizerman. They shared the paper at the 20th ACM International Conference on Mobile Systems, Applications, and Services.
