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

This paper describes a practical dual-process speech enhancement system thatadapts environment-sensitive frame-online beamforming (front-end) with helpfrom environment-free block-online source separation (back-end). To use minimumvariance distortionless response (MVDR) beamforming, one may train a deepneural network (DNN) that estimates time-frequency masks used for computing thecovariance matrices of sources (speech and noise). Backpropagation-basedrun-time adaptation of the DNN was proposed for dealing with the mismatchedtraining-test conditions. Instead, one may try to directly estimate the sourcecovariance matrices with a state-of-the-art blind source separation methodcalled fast multichannel non-negative matrix factorization (FastMNMF). Inpractice, however, neither the DNN nor the FastMNMF can be updated in aframe-online manner due to its computationally-expensive iterative nature. OurDNN-free system leverages the posteriors of the latest source spectrogramsgiven by block-online FastMNMF to derive the current source covariance matricesfor frame-online beamforming. The evaluation shows that our frame-online systemcan quickly respond to scene changes caused by interfering speaker movementsand outperformed an existing block-online system with DNN-based beamforming by5.0 points in terms of the word error rate.

DNN-Free Low-Latency Adaptive Speech Enhancement Based on Frame-Online Beamforming Powered by Block-Online FastMNMF