Eliza (Z.-H.) Michalopoulou

Matched field processing (MFP) is often employed to address source localization problems in the ocean. MFP relies on the calculation of replica fields for likely source locations, which are then correlated to the true sound field. The source location estimates are those location values that maximize the correlation. In the same manner, MFP can be also used to estimate parameters other than location that affect underwater sound propagation. In practice, since the propagation medium is usually only approximately known, MFP can be used for the simultaneous estimation of source location and environmental parameters. The estimation procedure, thus, becomes multi-dimensional, and exhaustive implementation is computationally inefficient. Instead, global optimization methods are employed in order to search efficiently the objective function (measure of correlation between data and replicas). In this talk we present an overview of MFP, highlighting the concepts and showing real data results. We then present a novel optimization method, TABU, that navigates efficiently on matched field surfaces. Through a comparison to fast simulated annealing, we show the potential of TABU in matched field inversion and discuss the additional advantages of TABU searches using principal components.