Precise and efficient localization of room reflections using compact microphone arrays

Abstract

Recent works have shown that precise localization of dominant reflections in acoustic environments can be achieved even by relatively small arrays if advanced beamforming concepts are employed. To this end subspace-based and steered beamformers-based localization techniques are implemented in the so-called eigenbeam domain leading to eigenbeam (EB)-ESPRIT, EB-MUSIC, and EB-Minimum Variance Distortionless Response (EB-MVDR) beamforming. The resulting algorithms can be deployed with small off-the-shelf spherical microphone arrays and can then lead to acoustic maps using only a single multichannel recording. They can furthermore be used as a tool for efficiently inferring the geometry of a room. In this contribution we will present a comparison of several methods in real-world scenarios and illustrate their distinctive properties, including robustness issues and known limitations.