Design and evaluation of a higher-order spherical microphone/ambisonic sound reproduction system for the acoustical assessment of concert halls

Abstract

Previous studies of the perception of concert hall acoustics have generally employed two methods for soliciting listeners' judgments. One method is to have listeners rate the sound in a hall while physically present in that hall. The other method is to make recordings of different halls and seat positions, and then recreate the environment for listeners in a laboratory setting via loudspeakers or headphones. In situ evaluations offer a completely faithful rendering of all aspects of the concert hall experience. However, many variables cannot be controlled and the short duration of auditory memory precludes an objective comparison of different spaces. Simulation studies allow for more control over various aspects of the evaluations, as well as A/B comparisons of different halls and seat positions. The drawback is that all simulation methods suffer from limitations in the accuracy of reproduction. If the accuracy of the simulation system is improved, then the advantages of the simulation method can be retained, while mitigating its disadvantages.

Spherical microphone array technology has received growing interest in the acoustics community in recent years for many applications including beamforming, source localization, and other forms of three-dimensional sound field analysis. These arrays can decompose a measured sound field into its spherical harmonic components, the spherical harmonics being a set of spatial basis functions on the sphere that are derived from solving the wave equation in spherical coordinates. Ambisonics is a system for two- and three-dimensional spatialized sound that is based on recreating a sound field from its spherical harmonic components. Because of these shared mathematical underpinnings, ambisonics provides a natural way to present fully spatialized renderings of recordings made with a spherical microphone array.

Many of the previously studied applications of spherical microphone arrays have used a narrow frequency range where the array's performance is optimal, whereas most naturally occurring auditory scenes are broadband. Ambisonic systems have been used extensively for presenting virtual auditory scenes, but not as extensively for measured scenes. The research presented here first examines the mechanics of presenting data obtained with a spherical microphone array via ambisonics. Binaural modeling and two perceptual studies are used to examine the relationship between the design of a spherical microphone array and perception of the playback, in terms of both auditory localization and other measures of sound quality. The final study examines presenting concert hall auralizations via this method, with a comparison to binaural rendering via headphones.