Chair of Research & Teaching in artificial intelligence

ADvanced underSea Intelligent Listening: ADSIL

This research group is one of 40 french AI Chair, delivered in 2020 to H. Glotin. It is focused on bioacoustics and ADvanced underSea Intelligent Listening (ADSIL). It aims to tackle Passive Acoustic Monitoring, Acoustic Recognition, Machine listening, Deep Learning, Inversion, Wavelet learning… It addresses distributed processing and IoT / network communications to survey biodiversity and complex soundscapes.

The Chair is divided in 3 main tasks :

T1) Direct Modelisation, Inverse Problem, HPC, Data Augmentation, Deep Learning

T2) Joint Classification & Localisation, Propagation of biophony or geophony

T3) Distributed Artificial Intelligence, network of Autonomous Vehicle, sonobuoy

Then they are combined in various paradigms :

A) Biosonar (T1, T2, T3)

  • Classification of biosonar
  • Matching pursuit and 3D tracking of biosonar
  • Study of the strategy of groups of biosonar hunters
  • Long survey of hunters
  • Bio-multistatism
  • Individual signatures
  • Ecosystem and Bioacoustics

B) Classification of voicings  (T1, T2) 

  • Individual call tracking, applied to Orca and others
  • Whale songs and evolution
  • Source separations, dialect
  • Individual signatures
  • HF and LF voiced pulse classification

C) AI & propagation, Explanable IA (T2, T3) .

  • Propagation & DNN / GPU RTX
  • Stream modelisation, multiscale forecasting

D) IA online, network IoT / Embedded IA  (T3) 

  • Online Bombyx2
  • Online KM3
  • Online Orcalab
  • Embedded AI
  • Ultra-low-power for listening

E) Maximisation of observation for classification & localisation (T1,T2, T3)  

  • Network of sensors (Bombyx sonobuoys, Sphyrnas ASV, KM3Env astrophysical observatory)
  • Network of fixed stations (Carimam, Orcalab)
  • Optimization with 3D acoustic serious games

This Project is granted by Agence nationale de la recherche and
Agence de l’innovation de défense