Assessing the Authenticity of Individual Dynamic Binaural Synthesis

Binaural technology allows to capture sound fields by recording the sound pressure arriving at the listener’s ear canal entrances. If these signals are reconstructed for the same listener the simulation should be indistinguishable from the corresponding real sound field. A simulation fulfilling this premise could be termed as perceptually authentic. Authenticity has been assessed previously for static binaural resynthesis of sound sources in anechoic environments, i.e. for HRTF-based simulations not accounting for head movements of the listeners. Results indicated that simulations were still discernable from real sound fields, at least, if critical audio material was used. However, for dynamic binaural synthesis to our knowledge – and probably because this technology is even more demanding – no such study has been conducted so far. Thus, having developed a state-of-the-art system for individual dynamic auralization of anechoic and reverberant acoustical environments, we assessed its perceptual authenticity by letting subjects directly compare binaural simulations and real sound fields. To this end, individual binaural room impulses were acquired for two different source positions in a medium-sized recording studio, as well as individual headphone transfer functions. Listening tests were conducted for two different audio contents applying a most sensitive ABX test paradigm. Results showed that for speech signals many of the subjects failed to reliably detect the simulation. For pink noise pulses, however, all subjects could distinguish the simulation from reality. Results further provided evidence for future improvements.DFG, WE 4057/3-1, Simulation and Evaluation of Acoustical Environments (SEACEN