Pruning masks from dynamic channel pruning in speech enhancement networks encode enough information for simple predictors to achieve 93% VAD accuracy, 84% noise classification accuracy, and R2=0.86 on F0 estimation with negligible overhead.
TGIF: Talker Group-Informed Familiariza- tion of Target Speaker Extraction
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
eess.AS 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
From Diet to Free Lunch: Estimating Auxiliary Signal Properties using Dynamic Pruning Masks in Speech Enhancement Networks
Pruning masks from dynamic channel pruning in speech enhancement networks encode enough information for simple predictors to achieve 93% VAD accuracy, 84% noise classification accuracy, and R2=0.86 on F0 estimation with negligible overhead.