ERes2NetV2: Boosting Short-Duration Speaker Verification Performance with Computational Efficiency
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2024
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866911903336366080 |
|---|---|
| author | Chen, Yafeng Zheng, Siqi Wang, Hui Cheng, Luyao Chen, Qian Zhang, Shiliang Li, Junjie |
| author_facet | Chen, Yafeng Zheng, Siqi Wang, Hui Cheng, Luyao Chen, Qian Zhang, Shiliang Li, Junjie |
| contents | Speaker verification systems experience significant performance degradation when tasked with short-duration trial recordings. To address this challenge, a multi-scale feature fusion approach has been proposed to effectively capture speaker characteristics from short utterances. Constrained by the model's size, a robust backbone Enhanced Res2Net (ERes2Net) combining global and local feature fusion demonstrates sub-optimal performance in short-duration speaker verification. To further improve the short-duration feature extraction capability of ERes2Net, we expand the channel dimension within each stage. However, this modification also increases the number of model parameters and computational complexity. To alleviate this problem, we propose an improved ERes2NetV2 by pruning redundant structures, ultimately reducing both the model parameters and its computational cost. A range of experiments conducted on the VoxCeleb datasets exhibits the superiority of ERes2NetV2, which achieves EER of 0.61% for the full-duration trial, 0.98% for the 3s-duration trial, and 1.48% for the 2s-duration trial on VoxCeleb1-O, respectively. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2406_02167 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | ERes2NetV2: Boosting Short-Duration Speaker Verification Performance with Computational Efficiency Chen, Yafeng Zheng, Siqi Wang, Hui Cheng, Luyao Chen, Qian Zhang, Shiliang Li, Junjie Audio and Speech Processing Signal Processing Speaker verification systems experience significant performance degradation when tasked with short-duration trial recordings. To address this challenge, a multi-scale feature fusion approach has been proposed to effectively capture speaker characteristics from short utterances. Constrained by the model's size, a robust backbone Enhanced Res2Net (ERes2Net) combining global and local feature fusion demonstrates sub-optimal performance in short-duration speaker verification. To further improve the short-duration feature extraction capability of ERes2Net, we expand the channel dimension within each stage. However, this modification also increases the number of model parameters and computational complexity. To alleviate this problem, we propose an improved ERes2NetV2 by pruning redundant structures, ultimately reducing both the model parameters and its computational cost. A range of experiments conducted on the VoxCeleb datasets exhibits the superiority of ERes2NetV2, which achieves EER of 0.61% for the full-duration trial, 0.98% for the 3s-duration trial, and 1.48% for the 2s-duration trial on VoxCeleb1-O, respectively. |
| title | ERes2NetV2: Boosting Short-Duration Speaker Verification Performance with Computational Efficiency |
| topic | Audio and Speech Processing Signal Processing |
| url | https://arxiv.org/abs/2406.02167 |