ICASSP 2021 Deep Noise Suppression Challenge: Decoupling Magnitude and Phase Optimization with a Two-Stage Deep Network
This work provides a significant improvement in speech denoising for real-time applications, benefiting users in telecommunications and voice interaction.
This paper addresses the challenge of speech signal recovery from noisy real-world acoustic environments. Their system achieved the top rank in the ICASSP 2021 Deep Noise Suppression Challenge for real-time track 1, based on Mean Opinion Score (MOS) with the ITU-T P.808 framework.
It remains a tough challenge to recover the speech signals contaminated by various noises under real acoustic environments. To this end, we propose a novel system for denoising in the complicated applications, which is mainly comprised of two pipelines, namely a two-stage network and a post-processing module. The first pipeline is proposed to decouple the optimization problem w:r:t: magnitude and phase, i.e., only the magnitude is estimated in the first stage and both of them are further refined in the second stage. The second pipeline aims to further suppress the remaining unnatural distorted noise, which is demonstrated to sufficiently improve the subjective quality. In the ICASSP 2021 Deep Noise Suppression (DNS) Challenge, our submitted system ranked top-1 for the real-time track 1 in terms of Mean Opinion Score (MOS) with ITU-T P.808 framework.