SiF-DeepVC Dataset

Fake Voice (AI-synthesized Voice) detection datasets based on speaker-irrelative features.

Description

If you need the fake voices generated by SiF-DeepVC, just use Part 3

• Part 1
  • "for-real-validation": original human recordings from FoR Validation dataset
  • "zh-real-test": original human recordings from MagicData Test dataset
• Part 2
  • "for-real-validation-denoised": slightly denoised "for-real-validation"
  • "zh-real-test-denoised": slightly denoised "zh-real-test"
  • "zh-real-test-silenced": silence-removed "zh-real-test"
• Part 3
  • "for-bh-madefake-final-r4k": cloned fake voices by SiF-DeepVC for Farid et al.
  • "for-deep4s-madefake-final-r4k": cloned fake voices by SiF-DeepVC for Deep4SNet
  • "for-rawnet-madefake-final-r4k": cloned fake voices by SiF-DeepVC for RawNet2

Download

Please note that the size of this zip file is about 7.8 GB.

URL: https://drive.google.com/file/d/1aaoG_jkREVBaglmW2M7cJmXU088Ndl8G/view?usp=sharing

Citation

We hope that you can cite our paper in your publication. It's not mandatory. :D

@INPROCEEDINGS{10301243,
  author={Liu, Xin and Tan, Yuan and Hai, Xuan and Yu, Qingchen and Zhou, Qingguo},
  booktitle={2023 IEEE 34th International Symposium on Software Reliability Engineering (ISSRE)}, 
  title={Hidden-in-Wave: A Novel Idea to Camouflage AI-Synthesized Voices Based on Speaker-Irrelative Features}, 
  year={2023},
  volume={},
  number={},
  pages={786-794},
  doi={10.1109/ISSRE59848.2023.00029}}