Adaptive Data Augmentation with NaturalSpeech3 for Far-field Speaker Verification
1 st Li Zhang, 2 nd Jiyao Liu, 3 nd Lei Xie · 2025
Read Adaptive Data Augmentation with NaturalSpeech3 for Far-field Speaker Verification by 1 st Li Zhang, 2 nd Jiyao Liu, 3 nd Lei Xie (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2501-08691.txt.
Description
Li Zhang, Jiyao Liu, and Lei Xie present a 2025 way to fake far off speech so voice ID systems can learn from it. Systems that check a speaker from across a room need much labeled far field speech, which is scarce. Adding close up speech helps little, since rooms change the sound.
Their method uses a strong speech making model to rebuild near speech with far room noise woven in. The voice traits come from the near sample, while the room traits come from real far samples. The fake samples keep who spoke but sound as if spoken far away. Tests aim to lift far field ID scores.
Catalog record
| CREATOR | 1 st Li Zhang, 2 nd Jiyao Liu, 3 nd Lei Xie |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 33 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2501-08691