6. 3D audio rendering through direct manipulation of in-situ recordings

The previous paragraphs in this dossier have focused on classic audio rendering approaches in which the sound environment is described as a set of point sources. Another possibility, which lends itself well to the capture and rendering of complex, realistic environments, is to render sound directly from spatial recordings. Unlike the classic case, in which each source emits a monophonic signal, input signals are generally multichannel and correspond to a spatial sampling of the sound environment. Like image-based graphics rendering techniques, which sample the plenoptic function [2] [54] [91], these techniques sample a "plenacoustic" function corresponding to the sound field. From this sampling, several applications are possible, such as spatial restitution, listening point interpolation or sound source displacement. In practice, however, the large number of channels required often limits...