Posterior Transition Modeling for Unsupervised Diffusion-Based Speech Enhancement
This work addresses speech enhancement for noisy audio processing, offering incremental improvements over existing diffusion-based methods.
The paper tackled unsupervised speech enhancement by proposing two diffusion-based algorithms that directly model conditional reverse transition distributions, eliminating hyperparameter tuning and providing exact likelihood scores, resulting in improved enhancement metrics and greater robustness to domain shifts on WSJ0-QUT and VoiceBank-DEMAND datasets.
We explore unsupervised speech enhancement using diffusion models as expressive generative priors for clean speech. Existing approaches guide the reverse diffusion process using noisy speech through an approximate, noise-perturbed likelihood score, combined with the unconditional score via a trade-off hyperparameter. In this work, we propose two alternative algorithms that directly model the conditional reverse transition distribution of diffusion states. The first method integrates the diffusion prior with the observation model in a principled way, removing the need for hyperparameter tuning. The second defines a diffusion process over the noisy speech itself, yielding a fully tractable and exact likelihood score. Experiments on the WSJ0-QUT and VoiceBank-DEMAND datasets demonstrate improved enhancement metrics and greater robustness to domain shifts compared to both supervised and unsupervised baselines.