Investigation of Typical Methods Applied to Compressive Matched Field Processing

Matched field processing (MFP) is widely studied in ocean acoustic localization. Its solution set is sparse since there are considerably fewer sources than replicas. Thus, we could potentially solve it as a sparse solution reconstruction problem. Compressed sensing (CS) provides the framework to achieve sparse solutions to a specific kind of under-determined problem showing excellent performance. In this paper, basis pursuit (BP) method and greedy method such as OMP are used to solve CS problem. We also use sparse Bayesian learning (SBL) to reconstruct the sparse solution via iterative algorithm. We study the effects of SNR, source amplitude, matrix coherence and number of snapshots on the performance of various algorithms and provide some insights about matrix coherence.