Article
On the use of real/virtual tests for the perceptual optimization of binaural rendering (en)
* Presenting author
Abstract:
Several techniques for the capture and rendering of binaural scenes exist. In particular, microphone array processing has proven to be particularly flexible, although to date, no method has been proven to be able to generate perceptually authentic percepts when compared to a real sound source in discrimination tests. In this contribution we present several pilot tests in which we compared binaurally generated sounds, rendered from measured Spatial Room Impulse Responses (SRIR), against real loudspeakers in the room, following both plausibility and transfer-plausibility paradigms. In these tests we investigated both spatial rendering as well as contextual aspects. Regarding rendering aspects, we investigated the required spatial resolution of early reflections, the frequency resolution of late reverberation, and the requirements for dynamic rendering of early reflections. The results suggest that all of the examined parameters can be simplified without significant perceptual effects. In terms of contextual variables, we evaluated the effects of listener expertise and stimuli on detection rates, concluding that both listener expertise and stimuli play an important role in the perceptual ratings. We will review the results of the experiments and potential avenues for perceptually motivated computational optimization of binaural rendering.