Array self-calibration using SAGE algorithm

Abstract

The performance of most existing array processing algorithms relies heavily on the precise knowledge of array manifold, which is decided by individual sensor characteristics and array configuration. A major challenge for self-calibration techniques is the increased computational burden due to additional perturbation parameters. In this contribution, a novel procedure for array self-calibration is presented. We apply the well known numerical method, the Space Alternating Generalized EM algorithm (SAGE), to simplify the multi-dimensional search procedure required for finding maximum likelihood (ML) estimates. Simulation shows that the proposed algorithm outperforms existing methods that are based on the small perturbation assumption. Furthermore, the proposed algorithm remain robust in critical scenarios including large sensor position errors and closely located signals.