The project is organized along the three following scientific challenges:
- The explicit modeling of speech production and perception.
- The statistical modeling of speech.
- The estimation and the exploitation of uncertainty in speech processing.
A large part of the research is conducted on French or English speech data; German and Arabic languages are considered for some tasks. Adaptation to other languages of the machine learning based approaches is possible, assuming the availability of corresponding corpora. Deep learning approaches are getting more and more used on many topics including
signal processing, speech recognition, and speech synthesis.
- Computer assisted learning
- Aided communication and monitoring
- Annotation and processing of spoken documents
- Multimodal computer interactions
MULTISPEECH is regularly involved in a number of collaborative projects with academic and/or industrial partners.
Academic and industrial relations
You can see a list of principal partners.
Here are the three last activity reports of the MULTISPEECH team.