Sound Spatializer
For all the ambient noise samples used in the dataset, the spatialization process was carried out using Spherebedev 3D sound spatialization sphere developed during Pierre Lecomte’s PhD in our lab, and the ambitools library, also developped by Pierre Lecomte during his PhD at Cnam.
The Spherebedev system is a spherical loudspeaker array with a radius of 1.07 meters, composed of 56 loudspeakers placed around the participants. To ensure precise spatialization in the full range of audio, two nested systems were used:
- A low-frequency system with 6 high-performance loudspeakers (ScanSpeak, up to 200 Hz) for accurate bass reproduction.
- A high-frequency system consisting of 50 loudspeakers (Aura, 2 inches, for frequencies above 200 Hz).
The multichannel audio used for higher order ambisonics resynthesis include third-order ambisonic recordings captured using a Zylia ZM-1S microphone, and a fifth-order ambisonic recordings captured with Memsbedev, a custom prototype ambisonic microphone built in our lab at LMSSC.