Higher order stereophony

Abstract

This thesis presents the derivation, experimental validation and perceptual evaluation of a new technique for spatial audio reproduction named Higher Order Stereophony. The approach uses the Taylor expansion of a plane wave soundfield to represent the field as a summation of its derivatives. The soundfield is then reproduced correctly along a single axis only, which is assumed to be the listener’s interaural axis, in an attempt to reproduce the correct binaural signals for 3D audio reproduction. A dynamic extension to the technique using head-tracking dynamically adapts the loudspeaker gains to ensure the reproduction axis always aligns with the listener’s interaural axis, regardless of the listener’s orientation.

Higher Order Stereophony is shown to be a generalisation to higher orders of classic stereo techniques such as the sine law, and is a frequency-independent solution resulting in loudspeaker panning gains. This generalisation is similar in manner to Higher Order Ambisonics, and parallels between the two are investigated including the derivation of decoders to transform from both the 3D or 2D Higher Order Ambisonics representation to Higher Order Stereophony, which first require a specific rotation of the soundfield followed by a matrix of gains. The new approach presents an alternative to Higher Order Ambisonics, with advantages in requiring only (N + 1) channels/loudspeakers for N-th order reproduction as well as requiring loudspeakers in front of the listener only.

Higher Order Stereophony is also applied to binaural rendering and shown to fully reproduce binaural signals with a rigid sphere Head-Related Transfer Function, due to the axisymmetric geometry of the scattering in the head model. When representing any Head-Related Transfer Function using spherical harmonics, a specific Higher Order Stereophony rotation is defined to align the interaural axis with the z axis, which is shown to reorder the energy of the spherical harmonic coefficients to those with m close to 0. Higher Order Stereophony is then demonstrated to reproduce all spherical harmonic coefficients of the Head-Related Transfer Function with m = 0 only. Subjective experiments comparing Higher Order Ambisonics and Higher Order Stereophony show that Higher Order Stereophony can perform similarly to Higher Order Ambisonics despite its advantages computationally and regarding loudspeaker positioning.

More information

Identifiers

ORCID for Jacob Hollebon: orcid.org/0000-0002-4119-4070

ORCID for Filippo Fazi: orcid.org/0000-0003-4129-1433

Catalogue record

Export record