Analysis of ABC Frontend Audio Systems for the NIST-SRE24
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , , , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866910959440756736 |
|---|---|
| author | Barahona, Sara Silnova, Anna Mošner, Ladislav Peng, Junyi Plchot, Oldřich Rohdin, Johan Zhang, Lin Han, Jiangyu Palka, Petr Landini, Federico Burget, Lukáš Stafylakis, Themos Cumani, Sandro Boboš, Dominik Hlavaček, Miroslav Kodovsky, Martin Pavlíček, Tomáš |
| author_facet | Barahona, Sara Silnova, Anna Mošner, Ladislav Peng, Junyi Plchot, Oldřich Rohdin, Johan Zhang, Lin Han, Jiangyu Palka, Petr Landini, Federico Burget, Lukáš Stafylakis, Themos Cumani, Sandro Boboš, Dominik Hlavaček, Miroslav Kodovsky, Martin Pavlíček, Tomáš |
| contents | We present a comprehensive analysis of the embedding extractors (frontends) developed by the ABC team for the audio track of NIST SRE 2024. We follow the two scenarios imposed by NIST: using only a provided set of telephone recordings for training (fixed) or adding publicly available data (open condition). Under these constraints, we develop the best possible speaker embedding extractors for the pre-dominant conversational telephone speech (CTS) domain. We explored architectures based on ResNet with different pooling mechanisms, recently introduced ReDimNet architecture, as well as a system based on the XLS-R model, which represents the family of large pre-trained self-supervised models. In open condition, we train on VoxBlink2 dataset, containing 110 thousand speakers across multiple languages. We observed a good performance and robustness of VoxBlink-trained models, and our experiments show practical recipes for developing state-of-the-art frontends for speaker recognition. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2505_15320 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Analysis of ABC Frontend Audio Systems for the NIST-SRE24 Barahona, Sara Silnova, Anna Mošner, Ladislav Peng, Junyi Plchot, Oldřich Rohdin, Johan Zhang, Lin Han, Jiangyu Palka, Petr Landini, Federico Burget, Lukáš Stafylakis, Themos Cumani, Sandro Boboš, Dominik Hlavaček, Miroslav Kodovsky, Martin Pavlíček, Tomáš Audio and Speech Processing Sound We present a comprehensive analysis of the embedding extractors (frontends) developed by the ABC team for the audio track of NIST SRE 2024. We follow the two scenarios imposed by NIST: using only a provided set of telephone recordings for training (fixed) or adding publicly available data (open condition). Under these constraints, we develop the best possible speaker embedding extractors for the pre-dominant conversational telephone speech (CTS) domain. We explored architectures based on ResNet with different pooling mechanisms, recently introduced ReDimNet architecture, as well as a system based on the XLS-R model, which represents the family of large pre-trained self-supervised models. In open condition, we train on VoxBlink2 dataset, containing 110 thousand speakers across multiple languages. We observed a good performance and robustness of VoxBlink-trained models, and our experiments show practical recipes for developing state-of-the-art frontends for speaker recognition. |
| title | Analysis of ABC Frontend Audio Systems for the NIST-SRE24 |
| topic | Audio and Speech Processing Sound |
| url | https://arxiv.org/abs/2505.15320 |