S2A Group: Audio Data Analysis and Signal Processing

Audio Data Analysis and Signal Processing

Link to the ADASP website

The theme ADASP: Audio Data Analysis & Signal Processing (formerly known as the AAO group) develops data analysis methods with a focus on sound data.

These developments are based on signal processing and statistical learning techniques, and focus mainly on methods for:

• Data decomposition and representation learning, in particular parsimonious representations,
• Parametric modelling.

These methods are used in tasks such as:

• Source separation,
• Description of content signals and scenes of human activity, including automatic classification,

and are applied to:

• MIR: Music Information Retrieval,
• Machine Listening,
• Analysis of multidimensional, heterogeneous or multimodal data, especially multimedia content,
• Musical acoustics,
• Analysis of physiological signals, especially electroencephalographic (EEG),
• Transformation of audio signals (denoising, enhancement, dereverberation, spatialization).

The team maintains close collaborations with other partners, both academic (Technical University of Berlin, Queen Mary University of London, Dublin University, ESPCI, IRCAM, INRIA-IRISA, INRIA-LORIA, CEA (Neurospin), INRIA-Parietal) and industrial (Orange, RTL, INA, Audionamix, Arkamys, Parrot…). These collaborations are partly developed in national or international projects.