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

A promising approach for multi-microphone speech separation involves two deepneural networks (DNN), where the predicted target speech from the first DNN isused to compute signal statistics for time-invariant minimum variancedistortionless response (MVDR) beamforming, and the MVDR result is then used asextra features for the second DNN to predict target speech. Previous studiessuggested that the MVDR result can provide complementary information for thesecond DNN to better predict target speech. However, on fixed-geometry arrays,both DNNs can take in, for example, the real and imaginary (RI) components ofthe multi-channel mixture as features to leverage the spatial and spectralinformation for enhancement. It is not explained clearly why the linear MVDRresult can be complementary and why it is still needed, considering that theDNNs and the beamformer use the same input, and the DNNs perform non-linearfiltering and could render the linear filtering of MVDR unnecessary. Similarly,in monaural cases, one can replace the MVDR beamformer with a monaural weightedprediction error (WPE) filter. Although the linear WPE filter and the DNNs usethe same mixture RI components as input, the WPE result is found tosignificantly improve the second DNN. This study provides a novel explanationfrom the perspective of the low-distortion nature of such algorithms, and findsthat they can consistently improve phase estimation. Equipped with thisunderstanding, we investigate several low-distortion target estimationalgorithms including several beamformers, WPE, forward convolutive prediction,and their combinations, and use their results as extra features to train thesecond network to achieve better enhancement. Evaluation results on single- andmulti-microphone speech dereverberation and enhancement tasks indicate theeffectiveness of the proposed approach, and the validity of the proposed view.

Leveraging Low-Distortion Target Estimates for Improved Speech Enhancement