Titre :

Design Choices for X-vector Based Speaker Anonymization

Auteur(s) :

Srivastava, Brij Mohan Lal [Auteur]
Machine Learning in Information Networks [MAGNET]
Tomashenko, Natalia [Auteur]
Laboratoire Informatique d'Avignon [LIA]
Wang, Xin [Auteur]
National Institute of Informatics [NII]
Vincent, Emmanuel [Auteur]
Speech Modeling for Facilitating Oral-Based Communication [MULTISPEECH]
Yamagishi, Junichi [Auteur]
National Institute of Informatics [NII]
Maouche, Mohamed [Auteur]
Distribution, Recherche d'Information et Mobilité [DRIM]
Bellet, Aurelien [Auteur]
Machine Learning in Information Networks [MAGNET]
Tommasi, Marc [Auteur]
Université de Lille

Titre de la manifestation scientifique :

INTERSPEECH 2020

Organisateur(s) de la manifestation scientifique :

International Speech Communication Association (ISCA)

Ville :

Shanghai

Pays :

Chine

Date de début de la manifestation scientifique :

2020-10-25

Mot(s)-clé(s) en anglais :

VoicePrivacy challenge
speaker anonymization
voice conversion
x-vectors
PLDA

Discipline(s) HAL :

Informatique [cs]
Informatique [cs]/Informatique et langage [cs.CL]
Informatique [cs]/Apprentissage [cs.LG]

Résumé en anglais : [en]

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 ...
Lire la suite >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.Lire moins >

Langue :

Anglais

Comité de lecture :

Oui

Audience :

Internationale

Vulgarisation :

Non

Projet ANR :

Apprentissage distribué, personnalisé, préservant la privacité pour le traitement de la parole

Collections :

• Centre de Recherche en Informatique, Signal et Automatique de Lille (CRIStAL) - UMR 9189

Source :

Harvested from HAL