Method of the main informative components for acoustic tomography

Abstract

Listen

Translate article

Techniques for an ocean acoustic tomography are based on the representation of the sound-speed field (SSF) by an expansion of some orthogonal basis functions. It is usually assumed that the most efficient basis functions for the SSF expansion are so-called empirical orthogonal functions (EOF). Meanwhile, the experiments show that algorithms based on the EOF can be extremely sensitive to slight errors in the assumed signal and noise characteristics. A new improved algorithm for acoustic tomography will be presented which minimizes output statistical errors of the SSF estimation. The concept of this algorithm is to use eigenvectors of the Fisher information operator as the basis functions into the SSF expansion. The algorithm involves a quadratic inequality constrain on the SSF expansion coefficients to select the most informative components among all the eigenvectors. It is shown that it leads to a robust algorithm in which statistical errors do not accumulate. The performance of the proposed algorithm will be illustrated both for deterministic and random signals.