Multi-Valued Eigen-Processing for Isolating Multiple Sources With a Rectangular Array

Michael D Collins,Joseph F Lingevitch

doi:10.1109/access.2021.3049777

Abstract

The multi-valued Bartlett (MVB) processor is useful for determining the locations of multiple acoustic sources in the ocean [J. Acoust. Soc. Am. 97, 235–241 (1995)]. This approach was originally applied to a vertical line array of hydrophones. The application to a rectangular array is explored here. The MVB processor is an eigen-processor that is based on the eigenvectors of the covariance matrix. It is multi-valued in the sense that an ambiguity surface is constructed for each member of a subset of the eigenvectors that correspond to the largest eigenvalues. The motivation for the approach is the fact that energy from different sources tends to partition into different eigenvectors. One of the advantages of the MVB processor on a rectangular array is that it is possible to determine if the partitioning is favorable without computing replica fields, which is often the most time-consuming task of matched-field processing computations. Examples are presented to illustrate the capabilities and limitations of the approach.

Highlights

A considerable amount of information may exist in a data set that is collected with an m × n rectangular array of acoustic receivers, where the horizontal dimension m and vertical dimension n are both substantially greater than one
A signal-processing technique that is designed for rectangular arrays is considered here for a matched-field processing [1]–[3] scenario in which the array is placed in an ocean acoustic waveguide with one of its axes oriented vertically
Ambiguity surfaces are formed by beamforming on a subset of the eigenvectors of the covariance matrix that correspond to the largest eigenvalues

Summary

INTRODUCTION

A considerable amount of information may exist in a data set that is collected with an m × n rectangular array of acoustic receivers, where the horizontal dimension m and vertical dimension n are both substantially greater than one. After the signals from different sources have been isolated from each other into different eigenvectors, the MVB processor exploits this partitioning by analyzing each eigenvector separately. As multiple sources move through a waveguide, there may be ambiguous points at which the acoustic fields from two or more sources are correlated on the array. At such points, the partitioning will be unfavorable, and the MVB processor will. F. Lingevitch: Multi-Valued Eigen-Processing for Isolating Multiple Sources With a Rectangular Array. It is possible to determine when the partitioning is favorable (i.e., when the signals from different sources have been isolated from each other) without generating any replica fields. It can be a substantial advantage to avoid generating replica fields, a task that requires computations that may not be practical and environmental information (e.g., sound speed and bathymetry) that may not be available

THE MVB PROCESSOR

DISCUSSION