English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

The decoupling-style concept begins to ignite in the speech enhancement area,which decouples the original complex spectrum estimation task into multipleeasier sub-tasks i.e., magnitude-only recovery and the residual complexspectrum estimation)}, resulting in better performance and easierinterpretability. In this paper, we propose a dual-branch federative magnitudeand phase estimation framework, dubbed DBT-Net, for monaural speechenhancement, aiming at recovering the coarse- and fine-grained regions of theoverall spectrum in parallel. From the complementary perspective, the magnitudeestimation branch is designed to filter out dominant noise components in themagnitude domain, while the complex spectrum purification branch is elaboratelydesigned to inpaint the missing spectral details and implicitly estimate thephase information in the complex-valued spectral domain. To facilitate theinformation flow between each branch, interaction modules are introduced toleverage features learned from one branch, so as to suppress the undesiredparts and recover the missing components of the other branch. Instead ofadopting the conventional RNNs and temporal convolutional networks for sequencemodeling, we employ a novel attention-in-attention transformer-based networkwithin each branch for better feature learning. More specially, it is composedof several adaptive spectro-temporal attention transformer-based modules and anadaptive hierarchical attention module, aiming to capture long-termtime-frequency dependencies and further aggregate intermediate hierarchicalcontextual information. Comprehensive evaluations on the WSJ0-SI84 +DNS-Challenge and VoiceBank + DEMAND dataset demonstrate that the proposedapproach consistently outperforms previous advanced systems and yieldsstate-of-the-art performance in terms of speech quality and intelligibility.

DBT-Net: Dual-branch federative magnitude and phase estimation with attention-in-attention transformer for monaural speech enhancement