Headphone Auralization of Acoustic Spaces Recorded with Spherical Microphone Arrays

Abstract

Binaural auralizations using headphones have become increasingly pop-ular in virtual reality applications to create a realistic three dimensional sound scape. Conventional methods use dummy heads to capture the sig-nals that arise at the ears of a listener. Such recordings cannot be played back with head-tracking, which is a significant limitation. Spherical micro-phone arrays allow capture of the spatial structure of a sound field. It is possible to impose the acoustic properties of the human body onto such a sound field and to calculate the signals that would arise at the ears of a listener who is coincident with the microphone array. Rotations of the listener, and therefore head-tracking, are straightforward to achieve. This thesis investigates signal processing of sound fields captured with spherical microphone arrays in combination with spherical measurements of head related impulse responses. When implemented, such signal processing enables true three dimensional audio with possible head tracking for all rotations, without limitations on the number of simultaneous users. The thesis has three major parts; the theory behind sound field expan-sions to the spherical harmonics domain and the related signal process-ing, controlled simulations to quantify the e˙ects di˙erent signal processing chains have on the resulting binaural signals, and a user study to see the subjective e˙ects of the proposed signal processing chains. A magnitude-phase separated expansion method is proposed to reduce time domain errors due to discrete spatial sampling, which is evaluated both with simulations and in the user study. The proposed expansion method performs worse than previously used expansion methods for non-anechoic spaces, resulting in artificial echoes in the binaural sound signals.