SURFACE VESSEL ACOUSTIC SIGNAL DIRECTION OF ARRIVAL ESTIMATION BY VECTOR SENSOR PROCESSING WITH THE MAXIMUM EIGENGAP ESTIMATOR

Abstract

This work intends to improve Navy/Marine Corps reconnaissance capabilities for operations in near-shore–contested amphibious environments. We present a novel unsupervised approach to vessel localization, the maximum eigengap estimator, by analyzing the multi-directional signal channels from a single vector sensor connected to the cabled observatory operated by the Monterey Bay Aquarium Research Institute–Monterey Accelerated Research System (MBARI-MARS). This work postulates that the direction of arrival and frequency of a vessel source signal can be identified by optimizing the weights over the aggregated cross-power spectral density matrices of a vector sensor’s directional channels. We explore the accuracy of the maximum eigengap estimator’s performance with various upper and lower bounds for frequency selection, and by changing source distance and direction to the sensor, and against multi-source scenarios. The estimator shows agreement with the physics-informed Bellhop ray model and produces accurate estimations for direction of arrival for single-source estimations within a 12-kilometer range from the sensor. We also demonstrate the novelty of the algorithm to produce automatic selection of source frequency.