Inria Innovation Lab Voice Technologie : Studio Maia & Multispeech project-team
In July 2017, Inria and Studio Maia joined forces within an Inria Innovation Lab named 'Voice Technologie'. The goal of this collaboration is to develop a software suite aimed at sound professionals for speech processing in the multimedia creation chain.
Created over 20 years ago, classified 'innovative company' and supported by Bpifrance, the CNC (French National Centre for Cinema and the Moving Image) and the CAP DIGITAL cluster, Studio Maia has established itself as one of the specialists in sound processing thanks to solutions at the cutting edge of innovation. Its versatility and expertise mean that it is able to offer all of the activities linked to audiovisual or cinematographic sound post-production , thereby addressing all of the technological sound challenges necessary for the production of a film soundtrack.
Based at the Inria Nancy Grand - Est centre, the Multispeech team is specialised in research relating to speech and sound processing. It focuses in particular on the multisource (source separation, robust speech and speaker recognition), multilingual (learning of foreign languages) and multimodal (audiovisual speech synthesis) aspects.
Its research programme is structured around three areas:
- explicit speech modelling which exploits its physical dimension,
- statistical modelling which relies on machine learning methods such as Bayesian models (HMM-GMM) and deep neural networks (DNN),
- consideration of the uncertainties linked to the strong variability of the speech signal and to model imperfections.