Generalized inverse beamforming with optimized regularization strategy

Abstract

A promising recent development on acoustic source localization and source strength estimation is the generalized inverse beamforming, which is based on the microphone array cross-spectral matrix eigenstructure. This method presents several advantages over the conventional beamforming, including a higher accuracy on the source center localization and strength estimation even with distributed coherent sources. This paper aims to improve the strength estimation of the generalized inverse beamforming method with an automated regularization factor definition. Also in this work, a virtual target grid is introduced, and source mapping and strength estimation are obtained disregarding, as much as possible, the reflections influence. Two simple problems are used to compare the generalized inverse performance with fixed regularization factor to performance obtained using the optimized regularization strategy. Numerical and experimental data are used, and two other strength estimation methods are also evaluated for reference.