INTERSPEECH 2021 Deep Noise Suppression Challenge

Chandan K. A. Reddy, Harishchandra Dubey, Kazuhito Koishida, Arun Asokan Nair, Vishak Gopal, Ross Cutler, Sebastian Braun, Hannes Gamper, Robert Aichner, Sriram Srinivasan

Abstract: The Deep Noise Suppression (DNS) challenge was designed to unify the research efforts in the area of noise suppression targeted for human perception. We recently organized a DNS challenge special session at INTERSPEECH 2020 and ICASSP 2021. We open-sourced training and test datasets for the wideband scenario along with a subjective evaluation framework based on ITU-T standard P.808, which was used to evaluate participants of the challenge. Many researchers from academia and industry made significant contributions to push the field forward, yet even the best noise suppressor was far from achieving superior speech quality in challenging scenarios. In this version of the challenge organized at INTERSPEECH 2021, we expanded our training and test datasets to accommodate fullband scenarios and challenging test conditions. We used ITU-T P.835 to evaluate the challenge winners as it gives additional information about the quality of processed speech and residual noise. The two tracks in this challenge focused on real-time denoising for (i) wideband, and (ii) fullband scenarios. We also made available a reliable non-intrusive objective speech quality metric for wideband called DNSMOS for the participants to use during their development phase.