Design Choices for X-vector Based Speaker Anonymization

Brij Mohan Lal Srivastava (1) , Natalia Tomashenko (2) , Xin Wang (3) , Emmanuel Vincent (4) , Junichi Yamagishi (3) , Mohamed Maouche (5) , Aurélien Bellet (1) , Marc Tommasi (6)

1 MAGNET - Machine Learning in Information Networks
2 LIA - Laboratoire Informatique d'Avignon
3 NII - National Institute of Informatics
4 MULTISPEECH - Speech Modeling for Facilitating Oral-Based Communication
5 DRIM - Distribution, Recherche d'Information et Mobilité
6 Université de Lille

The recently proposed x-vector based anonymization scheme converts any input voice into that of a random pseudo-speaker. In this paper, we present a flexible pseudo-speaker selection technique as a baseline for the first VoicePrivacy Challenge. We explore several design choices for the distance metric between speakers, the region of x-vector space where the pseudo-speaker is picked, and gender selection. To assess the strength of anonymization achieved, we consider attackers using an x-vector based speaker verification system who may use original or anonymized speech for enrollment, depending on their knowledge of the anonymization scheme. The Equal Error Rate (EER) achieved by the attackers and the decoding Word Error Rate (WER) over anonymized data are reported as the measures of privacy and utility. Experiments are performed using datasets derived from LibriSpeech to find the optimal combination of design choices in terms of privacy and utility.

Mots clés: VoicePrivacy challenge; speaker anonymization; voice conversion; x-vectors; PLDA

Domaines: Informatique [cs]; Informatique et langage [cs.CL]; Apprentissage [cs.LG]

Brij Mohan Lal Srivastava, Natalia Tomashenko, Xin Wang, Emmanuel Vincent, Junichi Yamagishi, et al.. Design Choices for X-vector Based Speaker Anonymization. INTERSPEECH 2020, International Speech Communication Association (ISCA), Oct 2020, Shanghai, China. ⟨hal-02610447v2⟩