Research interests
acoustic modeling
end-to-end models
speaker adaptation
self-/semi- supervised learning
end-to-end models
self-/semi- supervised learning
end-to-end models
self-/semi- supervised learning
transfer learning
adaptation to historical semantic content
representation learning
evaluation methodology
privacy and utility metrics
voice anonymization methods
VoicePrivacy challenges
privacy in acoustic modeling
federated learning
Some (selected) recent projects
Esperanto (Exchanges for SPEech ReseArch aNd TechnOlogies)
https://cordis.europa.eu/project/id/101007666
End-to-End Neural Approaches for Speech Translation – ON-TRAC
https://on-trac.univ-avignon.fr/home ANR: https://anr.fr/Projet-ANR-18-CE23-0021
Open data, tools and challenges for speaker anonymization – Harpocrates
https://anr.fr/Project-ANR-19-DATA-0008
JST-ANR VoicePersonae project: Speaker identity cloning and protection
https://nii-yamagishilab.github.io/voicepersonae
LeBenchmark
Benchmarking speech self-supervised learning models: http://lebenchmark.com
SELMA
Shaping speech technologies for media monitoring: https://selma-project.eu/