RETRIEVING SPEAKER INFORMATION FROM PERSONALIZED ACOUSTIC MODELS FOR SPEECH RECOGNITION

Mdhaffar, Salima; Bonastre, Jean-François; Tommasi, Marc; Tomashenko, Natalia; Estève, Yannick

Document type :

Autre communication scientifique (congrès sans actes - poster - séminaire...): Communication dans un congrès avec actes

Title :

RETRIEVING SPEAKER INFORMATION FROM PERSONALIZED ACOUSTIC MODELS FOR SPEECH RECOGNITION

Author(s) :

Mdhaffar, Salima [Auteur]
Laboratoire Informatique d'Avignon [LIA]
Bonastre, Jean-François [Auteur]
Laboratoire Informatique d'Avignon [LIA]
Tommasi, Marc [Auteur]
Machine Learning in Information Networks [MAGNET]
Tomashenko, Natalia [Auteur]
Laboratoire Informatique d'Avignon [LIA]
Estève, Yannick [Auteur]
Laboratoire Informatique d'Avignon [LIA]

Conference title :

IEEE ICASSP 2022

City :

Singapour

Country :

Singapour

Start date of the conference :

2022

English keyword(s) :

Automatic speech recognition
acoustic model
personalized acoustic models
collaborative learning
speaker information

HAL domain(s) :

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

English abstract : [en]

The widespread of powerful personal devices capable of collecting voice of their users has opened the opportunity to build speaker adapted speech recognition system (ASR) or to participate to collaborative learning of ASR. ...
Show more >The widespread of powerful personal devices capable of collecting voice of their users has opened the opportunity to build speaker adapted speech recognition system (ASR) or to participate to collaborative learning of ASR. In both cases, personalized acoustic models (AM), i.e. fine-tuned AM with specific speaker data, can be built. A question that naturally arises is whether the dissemination of personalized acoustic models can leak personal information. In this paper, we show that it is possible to retrieve the gender of the speaker, but also his identity, by just exploiting the weight matrix changes of a neural acoustic model locally adapted to this speaker. Incidentally we observe phenomena that may be useful towards explainability of deep neural networks in the context of speech processing. Gender can be identified almost surely using only the first layers and speaker verification performs well when using middle-up layers. Our experimental study on the TED-LIUM 3 dataset with HMM/TDNN models shows an accuracy of 95% for gender detection, and an Equal Error Rate of 9.07% for a speaker verification task by only exploiting the weights from personalized models that could be exchanged instead of user data.Show less >

Language :

Anglais

Peer reviewed article :

Oui

Audience :

Internationale

Popular science :

Non

ANR Project :

Apprentissage distribué, personnalisé, préservant la privacité pour le traitement de la parole
Clonage et protection de l'identité vocale

Collections :