Supplementary material to the paper The VoicePrivacy 2020 Challenge: Results and findings

Natalia Tomashenko (1) , Xin Wang (2) , Emmanuel Vincent (3) , Jose Patino (4) , Brij Mohan Lal Srivastava (5, 3) , Paul-Gauthier Noé (1) , Andreas Nautsch (4) , Nicholas Evans (4) , Junichi Yamagishi (2, 6) , Benjamin O'Brien (7) , Anaïs Chanclu (1) , Jean-François Bonastre (1) , Massimiliano Todisco (4) , Mohamed Maouche (5)

1 LIA - Laboratoire Informatique d'Avignon
2 NII - National Institute of Informatics
3 MULTISPEECH - Speech Modeling for Facilitating Oral-Based Communication
4 Eurecom [Sophia Antipolis]
5 MAGNET - Machine Learning in Information Networks
6 Edin. - University of Edinburgh
7 LPL - Laboratoire Parole et Langage

The VoicePrivacy 2020 Challenge focuses on developing anonymization solutions for speech technology. This report complements the summary results and analyses presented by Tomashenko et al. (2021). After quickly recalling the challenge design and the submitted anonymization systems, we provide more detailed results and analyses. First, we present objective evaluation results for the primary challenge metrics and for alternative metrics and attack models, and we compare them with each other. Second, we present subjective evaluation results for speaker verifiability, speech naturalness, and speech intelligibility. Finally, we compare these objective and subjective evaluation results with each other.

Keywords: automatic speech recognition; attack model; metrics; utility; privacy; anonymization; speech synthesis; voice conversion; speaker verification

Domains: Artificial Intelligence [cs.AI]

Natalia Tomashenko, Xin Wang, Emmanuel Vincent, Jose Patino, Brij Mohan Lal Srivastava, et al.. Supplementary material to the paper The VoicePrivacy 2020 Challenge: Results and findings. 2021. ⟨hal-03335126v2⟩