The VoicePrivacy 2022 Challenge Evaluation Plan

@article{Tomashenko2022TheV2,
  title={The VoicePrivacy 2022 Challenge Evaluation Plan},
  author={Natalia A. Tomashenko and Brij Mohan Lal Srivastava and Xin Wang and Pierre Champion and Emmanuel Vincent and Andreas Nautsch and Junichi Yamagishi and Nicholas W. D. Evans and Jean-François Bonastre and Paul-Gauthier No{\'e} and Massimiliano Todisco and Jose Patino},
  journal={ArXiv},
  year={2022},
  volume={abs/2205.07123}
}
The VoicePrivacy Challenge aims to promote the development of privacy preservation tools for speech technology by gathering a new community to define the tasks of interest and the evaluation methodology, and benchmarking solutions through a series of challenges. In this document, we formulate the voice anonymization task selected for the VoicePrivacy 2020 Challenge and describe the datasets used for system development and evaluation. We also present the attack models and the associated objective… 

Figures and Tables from this paper

Introducing the VoicePrivacy Initiative
TLDR
The voice anonymization task selected for the VoicePrivacy 2020 Challenge is formulated and the datasets used for system development and evaluation are described, including two anonymization baselines and objective evaluation results.
The VoicePrivacy 2020 Challenge: Results and findings
Supplementary material to the paper The VoicePrivacy 2020 Challenge: Results and findings
TLDR
The VoicePrivacy 2020 Challenge focuses on developing anonymization solutions for speech technology and objective evaluation results for speaker verifiability, speech naturalness, and speech intelligibility are presented.
Post-evaluation analysis for the VoicePrivacy 2020Challenge: Using anonymized speech data to train attackmodels and ASR
TLDR
The anonymized data collection for two different problems is investigated: (1) training more advanced attack models and (2) using it in downstream tasks, in particular, for ASR training.
Analysis of PingAn Submission in the VoicePrivacy 2020 Challenge
TLDR
This paper shows the post-evaluation analysis of the efforts in the VoicePrivacy 2020 Challenge on speaker verification, and can greatly reduce the Equal Error Rate (EER) by using the proposed methods.
Speaker Anonymization with Distribution-Preserving X-Vector Generation for the VoicePrivacy Challenge 2020
TLDR
A Distribution-Preserving Voice Anonymization technique is presented which can be easily integrated with others as the anonymization component of the system and removes the need to distribute a pool of speakers to use during the anonymizing.
Lightweight Voice Anonymization Based on Data-Driven Optimization of Cascaded Voice Modification Modules
TLDR
This paper uses signal processing methods with training data for optimizing hyperparameters in a data-driven manner and succeeded in deteriorating the speaker recognition rate and improving the speech recognition rate compared to a signal processing-based conventional method.
Exploring the Importance of F0 Trajectories for Speaker Anonymization using X-vectors and Neural Waveform Models
TLDR
Modifying the F0 can improve speaker anonymization by as much as 8% with minor word-error rate degradation, according to the VoicePrivacy Challenge 2020 framework and datasets developed and evaluated.
Design Choices for X-vector Based Speaker Anonymization
TLDR
A flexible pseudo-speaker selection technique is presented as a baseline for the first VoicePrivacy Challenge and several design choices for the distance metric between speakers, the region of x-vector space where the pseudo- Speaker is picked, and gender selection are explored.
Speaker anonymisation using the McAdams coefficient
TLDR
Results derived using common VoicePrivacy 2020 databases and protocols show that random, optimised transformations can outperform competing solutions in terms of anonymisation while causing only modest, additional degradations to intelligibility, even in the case of a semi-informed privacy adversary.
...
...

References

SHOWING 1-10 OF 50 REFERENCES
Introducing the VoicePrivacy Initiative
TLDR
The voice anonymization task selected for the VoicePrivacy 2020 Challenge is formulated and the datasets used for system development and evaluation are described, including two anonymization baselines and objective evaluation results.
The VoicePrivacy 2020 Challenge: Results and findings
Supplementary material to the paper The VoicePrivacy 2020 Challenge: Results and findings
TLDR
The VoicePrivacy 2020 Challenge focuses on developing anonymization solutions for speech technology and objective evaluation results for speaker verifiability, speech naturalness, and speech intelligibility are presented.
Post-evaluation analysis for the VoicePrivacy 2020Challenge: Using anonymized speech data to train attackmodels and ASR
TLDR
The anonymized data collection for two different problems is investigated: (1) training more advanced attack models and (2) using it in downstream tasks, in particular, for ASR training.
Privacy and Utility of X-Vector Based Speaker Anonymization
We study the scenario where individuals (speakers) contribute to the publication of an anonymized speech corpus. Data users leverage this public corpus for downstream tasks, e.g., training an
Hidebehind: Enjoy Voice Input with Voiceprint Unclonability and Anonymity
TLDR
The evaluation substantiates the efficacy of VoiceMask, e.g., it is able to reduce the chance of a user's voice being identified from 50 people by a mean of 84%, while reducing voice input accuracy no more than 14.2%.
Towards Privacy-Preserving Speech Data Publishing
TLDR
A heuristic algorithm is proposed to personalize the sanitization for speakers to restrict their privacy leak (p-leak limit) while minimizing the utility loss, and a novel method - key term perturbation for speech contentsanitization is designed.
Design Choices for X-vector Based Speaker Anonymization
TLDR
A flexible pseudo-speaker selection technique is presented as a baseline for the first VoicePrivacy Challenge and several design choices for the distance metric between speakers, the region of x-vector space where the pseudo- Speaker is picked, and gender selection are explored.
Benchmarking and challenges in security and privacy for voice biometrics
TLDR
The ASVspoof challenge relating to the development of spoofing countermeasures; the VoicePrivacy initiative which promotes research in anonymisation for privacy preservation; and the benchmarking methodology that has spearheaded progress in traditional research are described.
Speaker anonymisation using the McAdams coefficient
TLDR
Results derived using common VoicePrivacy 2020 databases and protocols show that random, optimised transformations can outperform competing solutions in terms of anonymisation while causing only modest, additional degradations to intelligibility, even in the case of a semi-informed privacy adversary.
...
...