DS0703 - Le numérique au service des arts, du patrimoine, des industries culturelles et éditoriales

Online unmixing for large historical archives – KAMoulox

Submission summary

The 36 months KAMoulox project aims at developing cutting-edge audio restoration and separation techniques and at making them available for research, educational and artistic purpose to a large audience, in an open-source environment embedded in real-world sound archives. The main assumption of the project is that audio separation and restoration can help shape new practices regarding our immaterial audio heritage. It can first greatly benefit to the scientific and educational exploitation of this heritage through the concept of crowd processing, but it also provides the tools that turn our rare audio treasures into new raw material that may be manipulated and exploited for educational reuse or artistic creation.

The project comprises both an important theoretical research effort in probabilistic signal processing and an application-oriented methodology that targets the actual transfer of cutting-edge signal processing research to a wide audience. Recent breakthroughs in audio demixing have indeed made it possible to design new original methods and algorithms that can be used by the neophyte for the purpose of restoring and/or editing historical audio recordings easily.

From a scientific perspective, KAMoulox will be concerned with exploring new robust probabilistic models for multivariate time series, exploiting two promising research directions I have proposed recently. The scientific innovations leveraged by the project have already been shown as extremely promising and effective in international evaluation campaigns.

From an applicative perspective, KAMoulox will capitalize over an already existing baseline audio processing server to incrementally implement and release a client-server architecture that can be embarked into large historical audio repositories. All software developed within the project will be released publicly under an open Affero-GPL license. A dedicated KAMoulox budget is allocated to fund specialized professional software engineers to assist the researchers in this development effort.

Finally, KAMoulox comprises an important science outreach objective. First, the online audio editing application will be designed in close collaboration with a representative panel of the regular users of the large French Archive du CNRS --- Musée de l'Homme repository, who mostly are researchers in digital humanities. Second, pedagogical science outreach content will be created both to promote applied mathematics to a wide audience, and to raise the attention of the digital humanities community on the new possibilities yielded by recent audio processing research.

Project coordinator

Monsieur Antoine Liutkus (Institut National de Recherche en Informatique et en Automatique)

The author of this summary is the project coordinator, who is responsible for the content of this summary. The ANR declines any responsibility as for its contents.


INRIA Institut National de Recherche en Informatique et en Automatique

Help of the ANR 282,828 euros
Beginning and duration of the scientific project: September 2015 - 36 Months

Useful links

Explorez notre base de projets financés



ANR makes available its datasets on funded projects, click here to find more.

Sign up for the latest news:
Subscribe to our newsletter