For joint denoising and dereverberation with diffusion models, cascade quality depends strongly on which distortion is removed first, and a single mixed-objective model gives the best overall compromise.
Deep neural network techniques for monaural speech enhancement and separation: state of the art anal- ysis,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
eess.AS 1years
2025 1verdicts
CONDITIONAL 1roles
background 1polarities
support 1representative citing papers
citing papers explorer
-
On the Application of Diffusion Models for Simultaneous Denoising and Dereverberation
For joint denoising and dereverberation with diffusion models, cascade quality depends strongly on which distortion is removed first, and a single mixed-objective model gives the best overall compromise.