Mask polarization restores bimodality in SE model predictions via Wasserstein distance at test time, delivering consistent gains across domain shifts and architectures.
Datasets All models were trained on the source dataset EARS-WHAM! (EARS-W) [ 9, 10] and evaluated on the 9 target datasets proposed by [ 1] to cover a wide range of domain shifts
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
-
Test-Time Adaptation For Speech Enhancement Via Mask Polarization
Mask polarization restores bimodality in SE model predictions via Wasserstein distance at test time, delivering consistent gains across domain shifts and architectures.