Partenaires

Octogone
Logo CNRS Logo UT2J
Logo MSHS-T



Rechercher

Sur ce site


Accueil du site > Actualités > Call of project ANR Corpus - CIESS Project

Call of project ANR Corpus - CIESS Project

par admin - 20 août 2012

We have just selected for the project financing CIESS. This project was submitted by the Interdisciplinary Unit Research Octogone using PETRA (leader) in response to a call for projects ANR "corpus". Our partners are the team SAMOVA from IRIT Lab and the young private partner AudioGaming, from Toulouse.

The project CIESS (Constitution, Indexing and Enrichment of an environmental Sound Corpus) is an ambitious project. You can read below the summary submitted :

"The sound corpus of environmental sounds are necessary for scientific studies of human perception and the creation of immersive environments (simulation, virtual reality, video games, ...). The use of the corpus is limited by two types of problems. Firstly the difficulties to annotate sounds make the dissemination of corpus data between scientists very difficult. Secondly, the existing corpus does not always meet the needs of scientists.

The project CIESS "Constitution, indexing and Enrichment of a corpus of Sound Environment" aims to provide practical solutions for the creation, use, and adaptation of the corpus of environmental sounds. This ambitious project will create a strong interaction between three scientific fields : perception, analysis and sound synthesis. These three components, rarely together, are necessary to understand all the issues related to the creation and the use of such corpus. CIESS takes a systemic approach considering the entire problem and proposes an important innovation in the design points of enrichment technologies corpus. The project brings together three partners specializing respectively in each of these areas :
- PETRA, part of the Interdisciplinary Unit Research OCTOGONE, Toulouse 2 University ;
- SAMOVA Team, part of the large IRIT laboratory, Toulouse 3 University ;
- AUDIOGAMMING society, accompanied by the incubator Midi-Pyrénées and winner of the contest to create innovative company of the Ministry of Research.

The CIESS project aims is to create a corpus of annotated environmental sounds very available to the scientific community as well as manufacturers and made available through the Very Large Equipment (TGE) ADONIS.

On the other hand CIESS project will be the first to propose a model infrastructure for enriching Sound Corpus. This infrastructure uses analysis tools and automatic synthesis of sound created specifically for the project. The PETRA Team (OCTOGONE) coordinate this development in an iterative manner to ensure that the advanced tools available and meet the requirements of perceptual studies.

The SAMOVA Team (IRIT) will develop robust tools for automatic annotation. These tools will be adapted to strong constraints related to the study of sounds made in real life will be developed by the IRIT. They will firstly an annotation noise events, and secondly the identification of acoustic descriptors relevant to perceptual criteria studied.

AUDIOGAMMING company, specialized in the synthesis and sound design, ensure the enrichment part of the corpus. A functional model of infrastructure will be created and the creation of ad-hoc sound synthesis using annotations and automatic perceptual act and will directly modify the elements of sound corpus. The infrastructure will also rely on the recommendations of perceptual order established by the PETRA Team (OCTOGONE).

Evaluation stages are planned throughout the project to test the robustness and effectiveness of the developed tools. "

Corpuses of environmental sounds are heavily used for scientific studies on human perception as well as a raw material for the creation of immersive soundscape for sound artists (immersive universe, video games …). The use of such corpus faces two major drawbacks. On one hand the annotation being a very tricky task with very few standards, it is difficult for scientists to disseminate corpuses, exchange data with peers or even search for existing corpuses and for specific sounds. On the other hand, existing corpuses does not always meet user needs.

Project CIESS, "Creation, Indexing and Enrichment of a corpus of environmental Sounds" aims to provide practical solutions and guidelines for creating, using and adapting such corpus. This ambitious project will create a strong interaction between the fields of perception, analysis and sound synthesis. These three components, rarely associated, are indeed necessary to grasps the whole complexity behind the creation and the use of specific sound corpuses. CIESS has a systemic approach considering the problem in its entirety and brings together three partners specialized respectively in each of these areas :

· OCTOGNE (PETRA) laboratory · IRIT laboratory, · AUDIOGAMMING, a young innovative company supported by incubator Midi-Pyrénées and two-times laureate of French Ministry of research and industry innovation contest.

CIESS will produce two major deliverables :

On one hand a corpus of environmental sounds thoroughly annotated and available to scientific communities as well as industrials and made available through the “Grand Equipement” program (TGE) ADONIS.

On the other hand CIESS will deliver the very first mock-up of a corpus enrichment framework. This framework will make use of automatic sound analysis tools as well as sound synthesis tools created specifically for that purpose. The OCTOGONE-PETRA laboratory will coordinate this development through an iterative process, in order to meet the requirements of perceptual studies.

IRIT will develop robust tools adapted to the high constraints of sound captured in real-life uncontrolled conditions. These tools allow the annotation of sound events, as well as the identification and computation of acoustic descriptors relevant to perceptual criteria studied.

AUDIOGAMMING, specialized in sound design and sound synthesis, will create the framework and the synthesis tools for the corpus enrichment. This framework is intended to act and to directly modify/enrich the sound events using automatic and perceptual annotations. These enrichments will be based on perceptual criteria est