Contribution
Evaluating the Reconstruction of Spectral Elevation Cues in Interpolated Sparsely Measured BRIRs
* Presenting author
Abstract:
Dynamic binaural rendering for augmented or virtual reality (AR/VR) audio typically relies on binaural room impulse responses (BRIRs), which can either be measured or synthesized. While measurements allow for the acquisition of individual BRIRs, which is of interest e.g. for the accurate virtualization of immersive audio studios, the effort to measure BRIRs for many head orientations with sufficient spatial resolution for artifact-free rendering is quite high. This effort can be reduced by measuring BRIRs for only a few head orientations and interpolating between them. A recent study following this approach applied the SARITA method, a cross-correlation-based time-aligned interpolation, and showed good results even for very sparse BRIR sets. However, it only assessed yaw head orientations with sound sources in the horizontal plane, leaving open the question of whether SARITA can accurately reconstruct spectral elevation cues. The present work examines the method’s performance for elevated sound sources and includes pitch head orientations. Furthermore, the study extends the binaural rendering based on the interpolated BRIRs to six degrees of freedom (6DoF) by adjusting the direct sound to listener translations. The technical and perceptual evaluation assesses the BRIR interpolation and the 6DoF binaural rendering on the example of an immersive audio studio.