Noise-aware Speech Enhancement using Diffusion Probabilistic Model
This work addresses speech enhancement for real-world applications with varying noise conditions, but it is incremental as it builds upon existing diffusion models.
The paper tackles the problem of speech enhancement for unseen noises by proposing a noise-aware diffusion model that uses noise-specific information to guide the denoising process, achieving improvements on multiple models, especially for unseen noises on the VB-DEMAND dataset.
With recent advances of diffusion model, generative speech enhancement (SE) has attracted a surge of research interest due to its great potential for unseen testing noises. However, existing efforts mainly focus on inherent properties of clean speech, underexploiting the varying noise information in real world. In this paper, we propose a noise-aware speech enhancement (NASE) approach that extracts noise-specific information to guide the reverse process in diffusion model. Specifically, we design a noise classification (NC) model to produce acoustic embedding as a noise conditioner to guide the reverse denoising process. Meanwhile, a multi-task learning scheme is devised to jointly optimize SE and NC tasks to enhance the noise specificity of conditioner. NASE is shown to be a plug-and-play module that can be generalized to any diffusion SE models. Experiments on VB-DEMAND dataset show that NASE effectively improves multiple mainstream diffusion SE models, especially on unseen noises.