Towards the prediction of perceived room acoustical similarity

Understanding perceived room acoustical similarity is crucial to generating perceptually optimized audio rendering algorithms that maximize the perceived quality while minimizing the computational cost. In this paper we present a perceptual study in which listeners compare dynamic binaural renderings generated from spatial room impulse responses (SRIRs) obtained in several rooms and positions and are asked to identify whether they belong to the same space. The perceptual results, together with monaural room acoustical parameters, are used to generate a prediction model that estimates the perceived similarity of two SRIRs