Equipe

Multispeech

Speech Modeling for Facilitating Oral-Based Communication

Presentation

Speech is the scientific object of our group, with the objectives of better understanding how humans product and perceive speech, and of elaborating efficient interactions between human beings and machines via automatic speech processing.
A first theme consists in studying the links between articulatory features, acoustic cues and human perception. Within the case of language learning this helps to design acoustic feedback which can improve speech production.
Providing feedback often requires acoustic cues to be extracted from the speech signal. This consti- tutes a second activity which becomes important with the emergence of applications concerning speech pathologies and disorders. The construction of models, either acoustic for automatic speech recognition, linguistic for speech to speech translation, or articulatory for studying speech production, is the third activity of our group and plays a central role in our scientific approach.

Research activities

• Multimodal speech
• Speech recognition and annotation
• Speech to speech translation

Keywords

Automatic speech recognition, stochastic modeling, phonetics, multimodal speech process- ing, speech synthesis, foreign language learning, speech disorders, speech to speech translation, articu- latory modeling, speech analysis

French Version

Analyse, perception et reconnaissance automatique de la parole