Skip to Main content Skip to Navigation

Identification de sources acoustiques complexes en milieu réverbérant par grands réseaux de microphones

Abstract : Knowing the directivity pattern of an acoustic source is useful in many applications in acoustics. To experimentally estimate the spatial signature, it is common to deploy microphones partially or totally surrounding the source. The acoustic radiation is then captured in all possible directions. In this thesis, we discuss the development of a large-scale 3D microphone array. This array, named "MODO" ("Les Murs Ont Des Oreilles", or, "The Walls Have Ears"), is comprised of 1024 digital MEMS microphones, flush mounted on the walls and the ceiling of a typical shoe-box room. In order to localize the sources and identify their directivity pattern, we solve the associated inverse problem under block-sparsity constraints. The chosen method exploits the small number of sources inside the room, allowing a sparse representation of the measured sound field. We use the spherical harmonics formalism to efficiently describe the directivity of the sources and their individual contributions to the radiation pattern. The acoustic path is modelled via integration of room transfer functions, synthesized with the mirror microphone method. We validated the proposed characterization method \textit{in situ} by comparison with known directivity patterns, calibrated using a high order spherical microphone array in controlled conditions.
Complete list of metadata
Contributor : Abes Star :  Contact
Submitted on : Friday, December 11, 2020 - 1:03:26 AM
Last modification on : Saturday, December 12, 2020 - 3:58:45 AM
Long-term archiving on: : Friday, March 12, 2021 - 6:29:37 PM


Version validated by the jury (STAR)


  • HAL Id : tel-03053332, version 1


Hugo Demontis. Identification de sources acoustiques complexes en milieu réverbérant par grands réseaux de microphones. Acoustique [physics.class-ph]. Sorbonne Université, 2019. Français. ⟨NNT : 2019SORUS196⟩. ⟨tel-03053332⟩



Record views


Files downloads