A quality-filtered 700-hour subset beats the full 2,500-hour URGENT2025 training set on perceptual quality metrics for both discriminative and generative speech enhancement models.
Conv-TasNet: Surpassing Ideal Time– Frequency Magnitude Masking for Speech Separation,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
eess.AS 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Less is More: Data Curation Matters in Scaling Speech Enhancement
A quality-filtered 700-hour subset beats the full 2,500-hour URGENT2025 training set on perceptual quality metrics for both discriminative and generative speech enhancement models.