Speaker
Description
Convincing auralisation in extended reality (XR) relies less on physical accuracy and more on perceptual plausibility, where a listener perceives a simulation as equivalent to a real acoustic event. Room impulse response (RIR) synthesis has been widely investigated, yet the relationship between numerical accuracy and perceptual plausibility remains comparatively underexplored, particularly when simulations rely on incomplete or inferred data. This work presents a perceptual comparison of reverberation synthesis paradigms spanning stochastic, parametric, geometric, wave-based and hybrid approaches. Rather than arbitrarily selecting implementations, the evaluated methods are chosen to represent distinct assumptions about how room acoustics can be modelled or approximated. These range from frequency-shaped stochastic decay models and compact parametric reverberators to explicit geometric simulations and scattering delay networks. Each method is provided with the corresponding scene information required for simulation, including material properties inferred semantically from the environment, such as wood flooring or glass surfaces.To isolate perceptual differences arising specifically from reverberation, all conditions share an identical direct-sound rendering pipeline, including individual Head Related Transfer Functions (HRTFs), distance attenuation and Direct-to-Reverberant Ratio (DRR) with only the reverberant field varying across conditions. Perceptual evaluation is performed through controlled 3 Degrees-of-Freedom (3DoF) listening tests using a two-alternative forced-choice (2AFC) paradigm, where participants select which of two renderings is perceived as more acoustically plausible for the presented room. In addition, a measured RIR is presented as one of the conditions to serve as a perceptual reference anchor. The proposed framework investigates how perceptual plausibility relates to reverberation structure, model complexity, and scene-description requirements across fundamentally different synthesis paradigms.