Investigating the Plausibility of Spatially Interpolated Binaural Room Impulse Responses

Rendering virtual acoustic environments with measured binaural room impulse responses (BRIRs) can provide a high level of realism. However, the measurement of BRIRs for multiple head orientations is typically tedious. Appropriate interpolation techniques can lower the demand for spatial resolution and thus reduce the measurement effort. A recent study compared several non-parametric interpolation strategies for very sparse multi-orientational BRIRs. The results suggest a high overall quality for BRIRs interpolated from sparse sets with angular resolutions of 30? and 40?. The present study investigates the realism of such sparsely measured BRIR sets for mixed reality (XR) applications compared to densely measured BRIRs in a plausibility experiment. The participants had to decide if a presented sound source was a real loudspeaker or a headphone simulation, only using their listening expectations as a inner reference. The results indicate that participants could not reliably identify the simulations for either dense or interpolated BRIRs. For dense BRIRs, participants responses were essentially random, suggesting they were guessing. For interpolated BRIRs, results indicate that participants were able to detect small sensory differences.