Improvements in the Perceived Quality of Streaming and Binaural Rendering of Ambisonics

With the increasing popularity of spatial audio content streaming and interactive binaural audio rendering, it is pertinent to study the quality of the critical components of such systems. This includes low-bitrate compression of Ambisonic scenes and binaural rendering schemes. This thesis presents a group of perceptual experiments focusing on these two elements of the Ambisonic delivery chain.

The first group of experiments focused on the quality of low-bitrate compression of Ambisonics. The first study evaluated the perceived timbral quality degradation introduced by the Opus audio codec at different bitrate settings and Ambisonic orders. This experiment was conducted using multi-loudspeaker reproduction as well as binaural rendering. The second study has been dedicated to auditory localisation performance in bitrate-compressed Ambisonic scenes reproduced over loudspeakers and binaurally using generic and individually measured HRTF sets. Finally, the third study extended the evaluated set of codec parameters by testing different channel mappings and various audio stimuli contexts. This study was conducted in VR thanks to a purposely developed listening test framework. The comprehensive evaluation of the Opus codec led to a set of recommendations regarding optimal codec parameters.

The second group of experiments focused on the evaluation of different methods for binaural rendering of Ambisonics. The first study in this group focused on the implementation of the established methods for designing Ambisonic-to-binaural filters and subsequent objective and subjective evaluations of these. The second study explored the concept of hybrid binaural rendering combining anechoic filters with reverberant ones. Finally, addressing the problem of non-individual HRTFs used for spatial audio rendering, an XR-based method for acquiring individual HRTFs using a single loudspeaker has been proposed.

The conducted perceptual evaluations identified key areas where the Ambisonic delivery chain could be improved to provide a more satisfactory user experience.