Predicting perceptual discriminability of room impulse responses at similar source–receiver distances using optimal transport

Quantifying the perceptual distance between room impulse responses (RIRs) is essential for evaluating room acoustic models in augmented reality/virtual reality (AR/VR) applications. While distances based on late-reverberation statistics are widely used, late-reverberation cues are typically weak for RIRs from the same room at similar source–receiver distances, whereas early-reflection timing differences remain. This letter studies the unbalanced optimal transport distance between RIR magnitude distributions as a predictor of perceptual discriminability in this regime. Evaluation on two independently collected discrimination datasets totaling 218 participants shows that the optimal transport distance correlates strongly with empirical discrimination rates.