Identification of Virtual Receiver Array Geometries that Minimize Audibility of Numerical Dispersion in Binaural Auralizations of Finite Difference Time Domain Simulations

摘要

This paper presents a perceptual evaluation of numerical dispersion in free-field headphone-based head-tracked binaural auralizations of finite difference time domain (FDTD) simulations. The simulated pressure, captured by virtual volumetric receiver arrays, is used to perform a spherical harmonics decomposition of the sound field and generate binaural signals. These binaural signals are compared perceptually to dispersion error-free binaural signals in a listening experiment designed using a duo-trio paradigm. The aim of the present work is to identify the size and density of the receiver array minimizing the audibility of numerical dispersion in the generated binaural signals. The spherical harmonics order was chosen to be 12 for the spatial decomposition. The overall reconstruction error, defined as the absolute value of the difference between the dispersion error-free and FDTD-simulated left-ear magnitude spectrum, was used as an objective metric to measure the spectral differences between the two signals. The listening experiment results show that this error does not correlate with the discrimination rates of the subjects. These results therefore suggest that this error does not suffice to describe the perceptual aspects introduced by numerical dispersion in the free-field dynamic binaural auralizations presented in the listening experiment. The results also show that increasing the receiver density for a fixed array size does not necessarily render numerical dispersion inaudible in the auralizations. Five out of 27 volumetric arrays led to FDTD-simulated binaural auralizations indistinguishable from the dispersion error-free binaural auralizations.