Capturing and reproducing realistic acoustic scenes for hearing research

Research output: Book/ReportPh.D. thesisResearch

678 Downloads (Pure)

Abstract

Accurate spatial audio recordings are important for a range of applications, from the creation of realistic virtual sound environments to the evaluation of communication devices, such as hearing instruments and mobile phones. Spherical microphone arrays are particularly well-suited for capturing spatial audio in three dimensions. However, practical constraints limit the number of microphones that can be used and thus the maximum spatial resolution and frequency bandwidth that can be achieved. Further, most important sound sources are near the horizontal plane, where human spatial hearing is also most accurate. This thesis therefore investigated whether the horizontal performance of spherical microphone arrays could be improved (i) through an appropriate placement of a fixed number of transducers on the sphere, and (ii) by applying mixed-order ambisonics (MOA) processing. MOA combines higher-order ambisonics (HOA) with additional, horizontally oriented spherical harmonic functions of higher orders. Simulations of a MOA array, with a higher density of microphones near the equator, and an array with a nearly uniform distribution of microphones were compared in terms of spatial resolution and robustness. A MOA array was constructed, and some of the simulation results were validated with measurements. Results showed that for MOA, the spatial resolution was improved for horizontal sources at mid to high frequencies and the robustness to noise and measurement errors was similar to that of HOA. The properties of MOA microphone layouts and processing were investigated further by considering several order combinations. It was shown that the performance for horizontal vs. elevated sources can be adjusted by varying the order combination, but that a benefit of the higher horizontal orders can only be seen at mid to high frequencies as the need for regularization limits spatial directivity at lower frequencies. Finally, the MOA array was also evaluated in terms of sound field reconstruction error in a head-sized region. Results provided a physical validation of the functioning of the MOA microphone array and further showed that the MOA approach results in a somewhat larger “sweet area” for horizontal sources than for elevated sound sources. While the focus was on the technical evaluation of the developed MOA system, potential perceptual effects concerning MOA and microphone array recordings in general are also discussed. The system developed in this work provides new possibilities for the investigation of human perception in realistic and complex acoustic environments.
Original languageEnglish
PublisherTechnical University of Denmark, Department of Electrical Engineering
Number of pages118
Publication statusPublished - 2014
SeriesContributions to Hearing Research
Volume18

Fingerprint Dive into the research topics of 'Capturing and reproducing realistic acoustic scenes for hearing research'. Together they form a unique fingerprint.

Cite this