Matched-field source localization under multi-coherent modal group model via covariance matrix matching

As a typical model-based signal processing method, matched-field processing (MFP) depends on the precise model of ocean acoustic propagation and measurement process to estimate the source locations and/or ocean environmental parameters. Any mismatch between the mathematical model and the real information channel will degrade or even break down the performance. The waveguide space-time evolution, especially for shallow water environments, would cause the coherence of individual normal modes get lost. Thus, the multi-coherent modal group (MCMG) model is recently proposed to statistically model such phenomenon. Due to the multi-coherent property, the received signal covariance has a multi-rank structure, which needs a modified performance bound to reveal the fundamental limitation in parameter estimation and a novel signal estimator to handle the problem. Previously, matched-covariance estimator (MCE), which matches the multi-rank received signal covariance matrix with the modeled signal covariance matrix, has shown a robust estimation capability against statistical mismatch under white noise conditions. In this paper, we revise the classic Cramer-Rao bound (CRB) to predict the performance of the estimation problem under the MCMG model in mean-square error (MSE) sense and bring in MCE to handle the multirank signal estimation problem. Performance analysis of MCE is implemented along with minimum variance distortionless response (MVDR) for source localization in a typical shallow water environment chosen from the 2001 Asian Seas International Acoustic Experiment (ASIAEX).