Revisiting and Refining Lagunas' Beamforming for Acoustic Imaging

Abstract

Lagunas et al proposed an adaptive beamforming method in 1986. However, this method never receives attention: until 2024, the citation of this paper is only 66. Actually, the Lagunas’ beamforming failed to identify sound sources if the required covariance matrix of array measurements is derived from the most commonly used sample covariance matrix. However, this paper shows that the Lagunas’ beamforming could be quite powerful if the covariance matrix was refined. It is justified theoretically that the Lagunas’ beamforming asymptotically achieves a perfect spatial resolution with an exponential convergence rate as its order increases. The robustness of acoustic imaging can be restored by exploiting the spiked covariance structure via eigenvalue clipping. Experimental results demonstrate that the refined Lagunas beamforming can outperform other popular adaptive beamforming methods, such as Bartlett, Capon, Pisarenko, and MUSIC, in the sense of higher spatial resolution.