Synthetic Room Impulse Responses and Reverberant Fake Speech

This dataset is made available under the Creative Commons Attribution 4.0 International License (CC BY 4.0).  

For any inquiries, please email: contact @ hieuthi.com

Contents

This dataset consists of two components:

- Synthetic Room Impulse Responses (RIRs): Derived from the MIT RIR Survey Dataset released by James Traer and Josh H. McDermott

- Reverberant Fake Speech Dataset: Derived from the C1 subset of the ASVspoof 2021 DF dataset

Notes

- The T60 and DRR values of the synthetic RIRs were estimated and may be highly inaccurate.  
- Please refer to the original publication for more details: Room Impulse Responses Help Attackers to Evade Deep Fake Detection

Citation

If you use this dataset in your research, please cite the following paper:

@inproceedings{luong2024room,
  title={Room impulse responses help attackers to evade deep fake detection},
  author={Luong, Hieu-Thi and Truong, Duc-Tuan and Lee, Kong Aik and Chng, Eng Siong},
  booktitle={2024 IEEE Spoken Language Technology Workshop (SLT)},
  pages={623--629},
  year={2024},
  organization={IEEE}
}