PGUSE combines a predictive speech enhancer with a diffusion model, fusing their outputs and truncating the diffusion start to improve universal speech enhancement with low inference cost.
Towards efficient models for real-time deep noise suppression,
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
-
A Composite Predictive-Generative Approach to Monaural Universal Speech Enhancement
PGUSE combines a predictive speech enhancer with a diffusion model, fusing their outputs and truncating the diffusion start to improve universal speech enhancement with low inference cost.