Audio Data Analysis and Signal Processing

The ADASP team (Audio Data Analysis and Signal Processing, formerly known as the AAO group) develops digital signal processing methods with applications to audio, music, multimodal and biomedical signals. Its activities range from theoretical work on machine learning for signal processing, signal models and sparse representations to computational optimization of algorithms.

Work is conducted on both a methodological level to develop new sound representations and models especially for musical signals on their application to practical problems. In particular, the group is interested in Adaptive methods for high resolution sinusoidal components tracking, sparse representations, Non-Negative Matrix factorization or hierarchical models and on their application to practical problems such as automatic indexing, compression or M/EEG signal processing. Source separation also appears to be at the heart of this research group with contributions at the methodological level and with applications in nearly all the individual research themes.

Besides this methodological axis, the research tackled by the group can be organized in three main themes :

  1. Machine listening and audio source separation:  The objective of this theme is to improve the capability of machines to analyse and interpret complex audio situations by developing specific digital signal processing methods. This is the main research theme of the group.
  2. Audio and multimodal signal processing:    The objective of this theme is first to develop novel generic models and approaches for audio signal representation and compression and second to automatically process multimodal data streams (segmentation, structuring,…).
  3. Biomedical signal analysis:  is dedicated to the analysis of biomedical signals, in particular electroencephalographic (EEG) and magnetoencephalographic (MEG).

The group maintains tight links with a number of academic partners  (Technical University of Berlin, Queen Mary Univeristy of London, Dublin University, ESPCI, IRCAM, INRIA-IRISA, INRIA-LORIA, INRIA-LORIA, CEA (Neurospin), INRIA-Parietal) and industrial prtners (Orange, RTL, INA, Audionamix, Arkamys, Parrot,. . .). These collaborations are often developed within national and international collaborative projects 

For more information, please check

We also propose a set of tools, software and datasets for data analysis, especially audio signal processing. Check the following link: