Seminar by Enzo Tiffeneau

Acoustic scene description by acoustic imaging

This seminar, which reviews the work carried out during the first year of the thesis, presents methods for accurately describing the sound environment of a scene. This involves locating, identifying, quantifying and describing the trajectory of the various sources present. The use of acoustic imaging methods has the advantage, with the aid of microphone arrays, to locate and separate the various sound sources, offering a detailed polyphonic representation of the scene. Once extracted, the nature of the signals can be identified using sound recognition models, the state of the art of which is achieved today by artificial intelligence models based on neural networks.

This will therefore cover the methods developed or used for separating and localizing sound sources, as well as the solutions implemented for their identification, together with the first results obtained.