Few-step Adversarial Schrödinger Bridge for Generative Speech Enhancement
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2025
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866916772577280000 |
|---|---|
| author | Han, Seungu Lee, Sungho Lee, Juheon Lee, Kyogu |
| author_facet | Han, Seungu Lee, Sungho Lee, Juheon Lee, Kyogu |
| contents | Deep generative models have recently been employed for speech enhancement to generate perceptually valid clean speech on large-scale datasets. Several diffusion models have been proposed, and more recently, a tractable Schrödinger Bridge has been introduced to transport between the clean and noisy speech distributions. However, these models often suffer from an iterative reverse process and require a large number of sampling steps -- more than 50. Our investigation reveals that the performance of baseline models significantly degrades when the number of sampling steps is reduced, particularly under low-SNR conditions. We propose integrating Schrödinger Bridge with GANs to effectively mitigate this issue, achieving high-quality outputs on full-band datasets while substantially reducing the required sampling steps. Experimental results demonstrate that our proposed model outperforms existing baselines, even with a single inference step, in both denoising and dereverberation tasks. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2506_01460 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Few-step Adversarial Schrödinger Bridge for Generative Speech Enhancement Han, Seungu Lee, Sungho Lee, Juheon Lee, Kyogu Sound Audio and Speech Processing Deep generative models have recently been employed for speech enhancement to generate perceptually valid clean speech on large-scale datasets. Several diffusion models have been proposed, and more recently, a tractable Schrödinger Bridge has been introduced to transport between the clean and noisy speech distributions. However, these models often suffer from an iterative reverse process and require a large number of sampling steps -- more than 50. Our investigation reveals that the performance of baseline models significantly degrades when the number of sampling steps is reduced, particularly under low-SNR conditions. We propose integrating Schrödinger Bridge with GANs to effectively mitigate this issue, achieving high-quality outputs on full-band datasets while substantially reducing the required sampling steps. Experimental results demonstrate that our proposed model outperforms existing baselines, even with a single inference step, in both denoising and dereverberation tasks. |
| title | Few-step Adversarial Schrödinger Bridge for Generative Speech Enhancement |
| topic | Sound Audio and Speech Processing |
| url | https://arxiv.org/abs/2506.01460 |