Sites Inria

Version française

PAROLE Research team

Analysis, perception and recognition of speech

  • Leader : Yves Laprie
  • Research center(s) : CRI Nancy - Grand Est
  • Field :
  • Theme :
  • Partner(s) : CNRS,Université de Lorraine
  • Collaborator(s) : CNRS, INRIA

Team presentation

Our project aims at automatically processing speech signals to understand their meaning, or to analyze and enhance their acoustic structure. It comes within the perspective of building efficient vocal interfaces and requires works in the fields of analysis, perception and automatic recognition of speech.

Research themes

  • Speech analysis Our research concerns the analysis and perception of acoustic cues, the acoustic-to-articulatory inversion and speech analysis. This research gives rise to several current or future applications in vocal rehabilitation therapy, improvement of hearing aids and language learning.

  • Speech modeling for automatic recognition Our research concerns stochastic models (Hidden Markov Models, Graphical models and Stochastic trajectory models), multiband models, speaker adaptation, channel adaptation and language models, and entail several current or future applications in automatic speech recognition, automatic dictation and text-to-speech alignment.

Speech is a multidisciplinary research field and combines works in phonetics as well as in pattern recognition. The multi-disciplinarity turns out to be a decisive asset to address new research issues, particularly learning languages and multiband approaches, that require both skills in automatic speech recognition and skills in phonetics.

International and industrial relations

  • Our industrial policy favors contracts which fit in fairly well with our scientific objectives. We are involved in several cooperations with industrial companies which exploit automatic speech recognition, particularly Mic2 in the framework of a RNRT project, Syncmagic PROCOMA in the framework of a PRIAMM project on lipsync and Babel Technologies which retails our speech analysis software WinSnoori.
  • Furthermore, we are involved in the European project ISAEUS in the field of vocal rehabilitation and we are working with teachers of foreign languages from Nancy in the framework of a regional project on learning languages.

Keywords: Knowledge representation Knowledge bases Object knowledge models Classification Categorisation Types Constraints Tasks Problem solving Molecular biology Multiple standpoints Medicine diagnosis Granula