Gespeichert in:
| Hauptverfasser: | , , , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2025
|
| Schlagworte: | |
| Online-Zugang: | https://arxiv.org/abs/2512.20978 |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866909975336452096 |
|---|---|
| author | Li, Haoyang Zhuang, Xuyi Adnan, Azmat Ni, Ye Rao, Wei Gopal, Shreyas Chng, Eng Siong |
| author_facet | Li, Haoyang Zhuang, Xuyi Adnan, Azmat Ni, Ye Rao, Wei Gopal, Shreyas Chng, Eng Siong |
| contents | Language Model (LM)-based generative modeling has emerged as a promising direction for TSE, offering potential for improved generalization and high-fidelity speech. We present GenTSE, a two-stage decoder-only generative LM approach for TSE: Stage-1 predicts coarse semantic tokens, and Stage-2 generates fine acoustic tokens. Separating semantics and acoustics stabilizes decoding and yields more faithful, content-aligned target speech. Both stages use continuous SSL or codec embeddings, offering richer context than discretized-prompt methods. To reduce exposure bias, we employ a Frozen-LM Conditioning training strategy that conditions the LMs on predicted tokens from earlier checkpoints to reduce the gap between teacher-forcing training and autoregressive inference. We further employ DPO to better align outputs with human perceptual preferences. Experiments on Libri2Mix show that GenTSE surpasses previous LM-based systems in speech quality, intelligibility, and speaker consistency. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2512_20978 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | GenTSE: Enhancing Target Speaker Extraction via a Coarse-to-Fine Generative Language Model Li, Haoyang Zhuang, Xuyi Adnan, Azmat Ni, Ye Rao, Wei Gopal, Shreyas Chng, Eng Siong Audio and Speech Processing Artificial Intelligence Machine Learning Language Model (LM)-based generative modeling has emerged as a promising direction for TSE, offering potential for improved generalization and high-fidelity speech. We present GenTSE, a two-stage decoder-only generative LM approach for TSE: Stage-1 predicts coarse semantic tokens, and Stage-2 generates fine acoustic tokens. Separating semantics and acoustics stabilizes decoding and yields more faithful, content-aligned target speech. Both stages use continuous SSL or codec embeddings, offering richer context than discretized-prompt methods. To reduce exposure bias, we employ a Frozen-LM Conditioning training strategy that conditions the LMs on predicted tokens from earlier checkpoints to reduce the gap between teacher-forcing training and autoregressive inference. We further employ DPO to better align outputs with human perceptual preferences. Experiments on Libri2Mix show that GenTSE surpasses previous LM-based systems in speech quality, intelligibility, and speaker consistency. |
| title | GenTSE: Enhancing Target Speaker Extraction via a Coarse-to-Fine Generative Language Model |
| topic | Audio and Speech Processing Artificial Intelligence Machine Learning |
| url | https://arxiv.org/abs/2512.20978 |