|Thesis abstract: |
In the past few years, the expectations of quality in sound rendering have significantly grown. Research has focused on making the listening experience more immersive and realistic by focusing on acquisition, analysis, processing and rendering of acoustic wave fields. In particular, in the past two decades, two approaches have established themselves as reference solutions for this class of applications: the former based on sound field decomposition into spherical harmonics (High-Order Ambisonics), and the latter based on a modal sound field decomposition (Wave Field Analysis/Synthesis). More recently, a new class of approaches based on geometric sound field decomposition have proved their potential in terms of quality and computational efficiency, offering new perspectives in the direction of interactive sound field modelling/rendering. Similar geometric decompositions have been successfully used for defining new geometric representations of acoustic wave fields, which are based on the so-called plenacoustic function, which describes the sound field in a spatial point along a given direction, a definition that is closely related to plenoptic function, currently popular in the field of computer vision. In this Ph.D. program I intend to bring together the most interesting geometric representations of wave fields that are emerging today, into a unified perspective that encompasses geometric wave field decomposition, plenacoustic representations, and processing in the ray space, for the purpose of acoustic scene reconstruction and, therefore, enhanced listening experiences. The ultimate goal is to define a unified framework that is ready to accommodate these new analysis/rendering metaphors. This framework should be based on a common processing layer that generates a wave field representation that can be analysed and processed using results from the literature of computer vision (image-based rendering) and pattern analysis.