Source localization using constrained Kalman beamforming in spherical harmonics domain

Abstract

A constrained Kalman beamforming (CKB) method in spherical harmonics domain is proposed to localize multiple sources captured by spherical arrays. It scans the Euclidean norm of the weights in three-dimensional space to estimate source directions. It can improve the accuracy of sound source localization when steering vector mismatch exits. Moreover, it has the advantage of rapid convergence and online implementation. Compared with the desired minimum variance distortionless response (MVDR) and diagonal loading (DL) methods, simulation results show that the proposed algorithm has robustness to arbitrary steering vector mismatch and low number of snapshots.