StuPASE: Towards Low-Hallucination Studio-Quality Generative Speech Enhancement
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2026
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866911501419282432 |
|---|---|
| author | Rong, Xiaobin Gao, Jun Wang, Zheng Yesilbursa, Mansur Wojcicki, Kamil Lu, Jing |
| author_facet | Rong, Xiaobin Gao, Jun Wang, Zheng Yesilbursa, Mansur Wojcicki, Kamil Lu, Jing |
| contents | Achieving high perceptual quality without hallucination remains a challenge in generative speech enhancement (SE). A representative approach, PASE, is robust to hallucination but has limited perceptual quality under adverse conditions. We propose StuPASE, built upon PASE to achieve studio-level quality while retaining its low-hallucination property. First, we show that finetuning PASE with dry targets rather than targets containing simulated early reflections substantially improves dereverberation. Second, to address performance limitations under strong additive noise, we replace the GAN-based generative module in PASE with a flow-matching module, enabling studio-quality generation even under highly challenging conditions. Experiments demonstrate that StuPASE consistently produces perceptually high-quality speech while maintaining low hallucination, outperforming state-of-the-art SE methods. Audio demos are available at: https://xiaobin-rong.github.io/stupase_demo/. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2603_09234 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | StuPASE: Towards Low-Hallucination Studio-Quality Generative Speech Enhancement Rong, Xiaobin Gao, Jun Wang, Zheng Yesilbursa, Mansur Wojcicki, Kamil Lu, Jing Audio and Speech Processing Achieving high perceptual quality without hallucination remains a challenge in generative speech enhancement (SE). A representative approach, PASE, is robust to hallucination but has limited perceptual quality under adverse conditions. We propose StuPASE, built upon PASE to achieve studio-level quality while retaining its low-hallucination property. First, we show that finetuning PASE with dry targets rather than targets containing simulated early reflections substantially improves dereverberation. Second, to address performance limitations under strong additive noise, we replace the GAN-based generative module in PASE with a flow-matching module, enabling studio-quality generation even under highly challenging conditions. Experiments demonstrate that StuPASE consistently produces perceptually high-quality speech while maintaining low hallucination, outperforming state-of-the-art SE methods. Audio demos are available at: https://xiaobin-rong.github.io/stupase_demo/. |
| title | StuPASE: Towards Low-Hallucination Studio-Quality Generative Speech Enhancement |
| topic | Audio and Speech Processing |
| url | https://arxiv.org/abs/2603.09234 |