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

Personalized speech enhancement (PSE) models achieve promising resultscompared with unconditional speech enhancement models due to their ability toremove interfering speech in addition to background noise. Unlike unconditionalspeech enhancement, causal PSE models may occasionally remove the target speechby mistake. The PSE models also tend to leak interfering speech when the targetspeaker is silent for an extended period. We show that existing PSE methodssuffer from a trade-off between speech over-suppression and interferenceleakage by addressing one problem at the expense of the other. We propose a newPSE model training framework using cross-task knowledge distillation tomitigate this trade-off. Specifically, we utilize a personalized voice activitydetector (pVAD) during training to exclude the non-target speech frames thatare wrongly identified as containing the target speaker with hard or softclassification. This prevents the PSE model from being too aggressive whilestill allowing the model to learn to suppress the input speech when it islikely to be spoken by interfering speakers. Comprehensive evaluation resultsare presented, covering various PSE usage scenarios.

Breaking the trade-off in personalized speech enhancement with cross-task knowledge distillation