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

Single-channel speech enhancement (SE) is an important task in speechprocessing. A widely used framework combines an analysis/synthesis filterbankwith a mask prediction network, such as the Conv-TasNet architecture. In suchsystems, the denoising performance and computational efficiency are mainlyaffected by the structure of the mask prediction network. In this study, we aimto improve the sequential modeling ability of Conv-TasNet architectures byintegrating Conformer layers into a new mask prediction network. To make themodel computationally feasible, we extend the Conformer using linear complexityattention and stacked 1-D dilated depthwise convolution layers. We trained themodel on 3,396 hours of noisy speech data, and show that (i) the use of linearcomplexity attention avoids high computational complexity, and (ii) our modelachieves higher scale-invariant signal-to-noise ratio than the improvedtime-dilated convolution network (TDCN++), an extended version of Conv-TasNet.

DF-Conformer: Integrated architecture of Conv-TasNet and Conformer using linear complexity self-attention for speech enhancement