[2510.16834] Schrödinger Bridge Mamba for One-Step Speech Enhancement

Computer Science > Sound arXiv:2510.16834 (cs) [Submitted on 19 Oct 2025 (v1), last revised 5 Mar 2026 (this version, v2)] Title:Schrödinger Bridge Mamba for One-Step Speech Enhancement Authors:Jing Yang, Sirui Wang, Chao Wu, Lei Guo, Fan Fan View a PDF of the paper titled Schr\"odinger Bridge Mamba for One-Step Speech Enhancement, by Jing Yang and 4 other authors View PDF HTML (experimental) Abstract:We present Schrödinger Bridge Mamba (SBM), a novel model for efficient speech enhancement by integrating the Schrödinger Bridge (SB) training paradigm and the Mamba architecture. Experiments of joint denoising and dereverberation tasks demonstrate SBM outperforms strong generative and discriminative methods on multiple metrics with only one step of inference while achieving a competitive real-time factor for streaming feasibility. Ablation studies reveal that the SB paradigm consistently yields improved performance across diverse architectures over conventional mapping. Furthermore, Mamba exhibits a stronger performance under the SB paradigm compared to Multi-Head Self-Attention (MHSA) and Long Short-Term Memory (LSTM) backbones. These findings highlight the synergy between the Mamba architecture and the SB trajectory-based training, providing a high-quality solution for real-world speech enhancement. Demo page: this https URL Comments: Subjects: Sound (cs.SD); Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Audio and Speech Processing (eess.AS) Cite as: arXiv:2510...