U-DREAM: Unsupervised Dereverberation guided by a Reverberation Model

2025-07-17

AI Summary

Supervised dereverberation methods struggle in real-world scenarios due to the scarcity of paired clean-reverberant speech data. Method: This paper proposes an unsupervised learning framework requiring only reverberant speech and a prior acoustic model. Its core innovation is a Bayesian inference–based sequential learning strategy that jointly optimizes neural network estimation of both the clean signal and room acoustic parameters (e.g., RT60, dimensions) via a reverberation-matching loss. By eliminating reliance on clean-speech labels, it establishes a weakly supervised-to-unsupervised end-to-end training paradigm. Contribution/Results: Experiments demonstrate that the method significantly outperforms existing unsupervised baselines using merely 100 samples annotated with reverberation parameters. It achieves more robust, interpretable, and generalizable speech dereverberation under low-resource conditions, offering strong practical applicability.

Technology Category

• Machine Learning: Unsupervised & Self-Supervised Learning
• Natural Language Processing: Learning & Optimization for NLP
• Search and Optimization: Learning to Search

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs

Abstract

This paper explores the outcome of training state-ofthe-art dereverberation models with supervision settings ranging from weakly-supervised to fully unsupervised, relying solely on reverberant signals and an acoustic model for training. Most of the existing deep learning approaches typically require paired dry and reverberant data, which are difficult to obtain in practice. We develop instead a sequential learning strategy motivated by a bayesian formulation of the dereverberation problem, wherein acoustic parameters and dry signals are estimated from reverberant inputs using deep neural networks, guided by a reverberation matching loss. Our most data-efficient variant requires only 100 reverberation-parameter-labelled samples to outperform an unsupervised baseline, demonstrating the effectiveness and practicality of the proposed method in low-resource scenarios.

Problem

• Unsupervised dereverberation using reverberant signals only
• Eliminates need for paired dry and reverberant data
• Effective in low-resource scenarios with minimal labeled samples

Innovation

• Unsupervised training with reverberant signals
• Bayesian sequential learning strategy
• Reverberation matching loss guidance