Advancing a Wavelet-Based Spatial Audio Format

This work further develops the theory of Spherical Wavelet Format (SWF), a spatial audio format inspired by Ambisonics that makes use of Spherical Wavelets as a basis to decompose the soundfield. In particular, we have specified a version of SWF that implements a method to build an arbitrary wavelet representation on an arbitrary triangular mesh. We make use of a modified lifting scheme to optimize the interpolating scaling functions for optimal playback reproduction, and we have demonstrated its functionality and competitiveness with state-of-the-art spatial audio algorithms on a 7.1.4 layout. The resulting SWF specification is available for use and further research in an open source python library. The python library is flexible enough to support any layout, and includes presets for the original Octahedral mesh from the first publication of SWF, as well as the 7.0.4-based SWF format used for objective and subjective evaluation in this report, and a Spherical Wavelet Format that naturally interpolates between standard surround sound formats (11.1.8,9.1.6,7.1.4,etc.). The library is intended to be used with the trivial decoding from the coarsest level of mesh, but can also be decoded using other strategies from less coarse representations.