Perceptual Evaluation of Mitigation Approaches of Errors due to Spatial Undersampling in Binaural Renderings of Spherical Microphone Array Data

Abstract

High fidelity virtual acoustic environments (VAEs) have become of increasing interest in the context of immersive virtual and augmented reality applications, and have developed to a reputable field of research. To present a real-life sound scene as VAE to a listener, the spatial sound field has to be captured. Traditionally, spatial sound fields are sequentially measured by means of a dummy head. Using spherical microphone array recordings is a promising alternative to time-consuming dummy head measurements. This enables the possibility of dynamic real-time applications that involve recordings and reproductions of spatial sound scenes as for example live concerts or telephone conferences. However, the quality of the rendered VAEs is significantly limited by the density of the microphone distribution provided by the microphone array. A limited number of microphones leads to spatial undersampling of the surrounding sound field and creates audible artifacts in the VAE. This work investigates the errors due to spatial undersampling and state-of-the-art approaches to mitigate them. In a concluding listening experiment, the improvements that can be achieved with these approaches were perceptually evaluated. It was found that there are a few algorithms that significantly improve the binaural auralization of spherical microphone array data.