Localization and Categorization of Early Reflections for Estimating Acoustic Reflection Coefficients

Abstract

Knowledge of room acoustic parameters such as frequency- and direction-dependent reflection coefficients, room volume, or geometric characteristics is important for the mod-eling of acoustic environments, e. g. to improve the plausibility of immersive audio in mixed reality applications or to transfer a physical acoustic environment into a completely virtual one. This paper presents a method for detecting first-order reflections in three-dimensions of spatial room impulse responses recorded with a spherical microphone array. By using geometric relations, the estimated direction of arrival (DOA), and the time difference of arrival (TDOA), the order of the respective mirror sound source is determined and categorized to the individual walls of the room. The detected DOA and TDOA of the first-order mirror sound sources are used to estimate the frequency-dependent reflection coefficients of the respective walls using a null-steering beamformer directed to the estimated DOA. Analysis in terms of DOA and TDOA indicates an accurate estimation for simulated and measured data. The estimation of the reflection coefficients shows a relative error of 3.5 % between 500 Hz and 4 kHz for simulated data. Furthermore, experimental challenges are discussed, such as the evaluation of the reflection coefficient estimation in real acoustic environments.