Is Phase Really Needed for Weakly-Supervised Dereverberation ?

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Rodrigues, Marius, Bahrman, Louis, Badeau, Roland, Richard, Gaël
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917096987820032
author Rodrigues, Marius
Bahrman, Louis
Badeau, Roland
Richard, Gaël
author_facet Rodrigues, Marius
Bahrman, Louis
Badeau, Roland
Richard, Gaël
contents In unsupervised or weakly-supervised approaches for speech dereverberation, the target clean (dry) signals are considered to be unknown during training. In that context, evaluating to what extent information can be retrieved from the sole knowledge of reverberant (wet) speech becomes critical. This work investigates the role of the reverberant (wet) phase in the time-frequency domain. Based on Statistical Wave Field Theory, we show that late reverberation perturbs phase components with white, uniformly distributed noise, except at low frequencies. Consequently, the wet phase carries limited useful information and is not essential for weakly supervised dereverberation. To validate this finding, we train dereverberation models under a recent weak supervision framework and demonstrate that performance can be significantly improved by excluding the reverberant phase from the loss function.
format Preprint
id arxiv_https___arxiv_org_abs_2511_17346
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Is Phase Really Needed for Weakly-Supervised Dereverberation ?
Rodrigues, Marius
Bahrman, Louis
Badeau, Roland
Richard, Gaël
Sound
Artificial Intelligence
Signal Processing
Classical Physics
Machine Learning
In unsupervised or weakly-supervised approaches for speech dereverberation, the target clean (dry) signals are considered to be unknown during training. In that context, evaluating to what extent information can be retrieved from the sole knowledge of reverberant (wet) speech becomes critical. This work investigates the role of the reverberant (wet) phase in the time-frequency domain. Based on Statistical Wave Field Theory, we show that late reverberation perturbs phase components with white, uniformly distributed noise, except at low frequencies. Consequently, the wet phase carries limited useful information and is not essential for weakly supervised dereverberation. To validate this finding, we train dereverberation models under a recent weak supervision framework and demonstrate that performance can be significantly improved by excluding the reverberant phase from the loss function.
title Is Phase Really Needed for Weakly-Supervised Dereverberation ?
topic Sound
Artificial Intelligence
Signal Processing
Classical Physics
Machine Learning
url https://arxiv.org/abs/2511.17346