Au(di)Noise
 Au(di)Noise = Au (di) [o] + (di) Noise [ing] |
|---|
Low Noise Case
High Noise Case
During the online meetings and presentations, probably many of us keep the fan switched off if it's too noisy. Otherwise, the listeners at the opposite side would never hear our voice. But it's hot too! How long can you keep fan off? That's why we need Noise Suppression mechanism embedded in the systems that will eliminate even the loud background noises so that only the enhanced voice reaches to the other side. Au(di)Noise, which comes from Audi-o De-Noise-ing, is a hybrid approach to suppress background noise during speaking.
Applications: Audio Headsets, Headphones, Voice Communication Devices, etc.
Timeline: August, 2020 - October, 2020
Collaborator(s): Achal Nilhani, Dr. S. Dhabal
Awards & Honors:
-
Our submission secured Global Rank 14 in the Deep Noise Suppression Challenge (version 2) organized by Microsoft Corporation.
Theory behind Working: Whenever Au(di)Noise incorporates a hybrid approach consisting of a modified adaptive filter to eliminate the stationary noise signals and a deep filter model to deal with the non-stationary noises. Using Au(di)Noise, we have achieved overall Mean Opinion Score (MOS) of 2.90 with MOS = 3.01 for Tonal and MOS = 3.00 for English Speech as targeted signal. [Max. possible MOS = 5.00] [CODE]