ICASSP 2021 Deep Noise Suppression Challenge: Decoupling Magnitude and Phase Optimization with a Two-Stage Deep Network

Li, Andong; Liu, Wenzhe; Luo, Xiaoxue; Zheng, Chengshi; Li, Xiaodong

doi:10.48550/arxiv.2102.04198

Cited by 2 publications

(8 citation statements)

References 30 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…To capture the long-term temporal dependencies, we insert cascaded temporal convolutional modules (TCMs) [10] in the bottleneck. To decrease the parameters, we opt to the squeezed version [15,16], i.e., S-TCM, where the feature size is first compressed into 64 rather than 512 as the literature stated [10], followed by dilated convolutions. For each stage, we stack three groups of TCMs, each of which includes 6 S-TCMs with dilation rate d = {1, 2, 4, 8, 16, 32}.…”

Section: Network Configurationsmentioning

confidence: 99%

“…We compare the proposed framework with five advanced SE systems, namely GCRN [8], DCCRN [34], TSCN [16], AECNN [35], and Conv-TasNet [10]. GCRN, DCCRN, and TSCN are complex-domain based approaches, which aim to recover both magnitude and phase information simultaneously.…”

Section: Baselinesmentioning

confidence: 99%

“…We evaluate the average DNSMOS score of our approach on the released blind test set, which is shown in Figure 2. Besides, the scores of DCCRN and TSCN+PP [16] are also provided, which won the first and second challenges, respectively. Several observations can be made.…”

Section: Dnsmos Comparison With Previous Championsmentioning

confidence: 99%

“…Inspired by the curriculum learning concept [13], multistage SE begin to thrive in the recent several years [11,14,15,16,17]. Different from the previous one-step paradigm, the whole mapping process is decomposed into several sub-stages and the network is coerced to learn following either "from easy to hard" or "from hard to easy" training paradigm [18].…”

Section: Introductionmentioning

confidence: 99%

“…Generally, these artifacts are audible and noncontinuous, which heavily hinder the subjective quality. To ameliorate the issue, we take a post-processing module to remove the remaining distortion [16], which proves significant to improve the MOS.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

A Simultaneous Denoising and Dereverberation Framework with Target Decoupling

Li¹,

Liu²,

Luo³

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

Background noise and room reverberation are regarded as two major factors to degrade the subjective speech quality. In this paper, we propose an integrated framework to address simultaneous denoising and dereverberation under complicated scenario environments. It adopts a chain optimization strategy and designs four sub-stages accordingly. In the first two stages, we decouple the multi-task learning w.r.t. complex spectrum into magnitude and phase, and only implement noise and reverberation removal in the magnitude domain. Based on the estimated priors above, we further polish the spectrum in the third stage, where both magnitude and phase information are explicitly repaired with the residual learning. Due to the data mismatch and nonlinear effect of DNNs, the residual noise often exists in the DNN-processed spectrum. To resolve the problem, we adopt a light-weight algorithm as the post-processing module to capture and suppress the residual noise in the non-active regions. In the Interspeech 2021 Deep Noise Suppression (DNS) Challenge, our submitted system ranked top-1 for the real-time track in terms of Mean Opinion Score (MOS) with ITU-T P.835 framework.

show abstract

Section: Network Configurationsmentioning

confidence: 99%

Section: Baselinesmentioning

confidence: 99%

Section: Dnsmos Comparison With Previous Championsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

A Simultaneous Denoising and Dereverberation Framework with Target Decoupling

Li¹,

Liu²,

Luo³

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

A Two-stage Complex Network using Cycle-consistent Generative Adversarial Networks for Speech Enhancement

Wang

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

Cycle-consistent generative adversarial networks (CycleGAN) have shown their promising performance for speech enhancement (SE), while one intractable shortcoming of these CycleGAN-based SE systems is that the noise components propagate throughout the cycle and cannot be completely eliminated.Additionally, conventional CycleGAN-based SE systems only estimate the spectral magnitude, while the phase is unaltered. Motivated by the multi-stage learning concept, we propose a novel two-stage denoising system that combines a CycleGAN-based magnitude enhancing network and a subsequent complex spectral refining network in this paper. Specifically, in the first stage, a CycleGAN-based model is responsible for only estimating magnitude, which is subsequently coupled with the original noisy phase to obtain a coarsely enhanced complex spectrum. After that, the second stage is applied to further suppress the residual noise components and estimate the clean phase by a complex spectral mapping network, which is a pure complex-valued network composed of complex 2D convolution/deconvolution and complex temporal-frequency atten-

show abstract

ICASSP 2021 Deep Noise Suppression Challenge: Decoupling Magnitude and Phase Optimization with a Two-Stage Deep Network

Cited by 2 publications

References 30 publications

A Simultaneous Denoising and Dereverberation Framework with Target Decoupling

A Simultaneous Denoising and Dereverberation Framework with Target Decoupling

A Two-stage Complex Network using Cycle-consistent Generative Adversarial Networks for Speech Enhancement

Contact Info

Product

Resources

About