Diffusion-based Speech Enhancement with Schr\"odinger Bridge and Symmetric Noise Schedule

TL;DR

This paper introduces SBSE, a diffusion-based speech enhancement method utilizing Schr"odinger Bridge, which directly models the transition between noisy and clean speech, outperforming baselines especially in low SNR scenarios with fewer inference steps.

Contribution

The paper proposes a novel Schr"odinger Bridge-based diffusion model for speech enhancement, directly learning the transition between noisy and clean speech, and incorporates a two-stage system with ratio mask information for improved performance.

Findings

Outperforms baseline models in low SNR conditions

Achieves state-of-the-art speech enhancement results

Requires fewer inference steps for optimal performance

Abstract

Recently, diffusion-based generative models have demonstrated remarkable performance in speech enhancement tasks. However, these methods still encounter challenges, including the lack of structural information and poor performance in low Signal-to-Noise Ratio (SNR) scenarios. To overcome these challenges, we propose the Schr\"oodinger Bridge-based Speech Enhancement (SBSE) method, which learns the diffusion processes directly between the noisy input and the clean distribution, unlike conventional diffusion-based speech enhancement systems that learn data to Gaussian distributions. To enhance performance in extremely noisy conditions, we introduce a two-stage system incorporating ratio mask information into the diffusion-based generative model. Our experimental results show that our proposed SBSE method outperforms all the baseline models and achieves state-of-the-art performance, especially in low SNR conditions. Importantly, only a few inference steps are required to achieve the best result.