A Two-Dimensional Threshold Test for Reverberation Time and Direct-to-Reverberant Ratio

In audio for augmented reality (AR), virtual sound sources should be seamlessly incorporated into a real acoustic environment, exhibiting acoustic properties that match those of the real acoustic space. One way of assessing the closeness of this match is to compute differences in acoustical parameters between measurements and renderings and to compare them to known just noticeable differences (JNDs) obtained from perceptual tests. However, JNDs for even the most common acoustical parameters are not well established and parameters that can be independently varied in artificial reverberators might not be perceptually independent. In this paper, we first demonstrate a method of exploring such perceptual dependence of room acoustic parameters by conducting a multidimensional threshold test, varying two parameters at the same time. The method is based on an adaptive 2D threshold estimation scheme, incorporating principles of active learning, which are well known in the machine learning community. We present first results of a specific test, varying reverberation time (RT) and direct-to-reverberant ratio (DRR) of a spatial room impulse response (SRIR) rendered dynamically to headphones. The test used a transferring task, where detectability is assessed using different source signals – arguably the most relevant design when results are to be considered in the context of forthcoming AR applications. We discuss the necessity and the challenges regarding multidimensional threshold tests in room acoustics and present the first indication showing that RT and DRR may not be perceptually independent.