The Maximal Eigengap Estimator for Acoustic Vector-Sensor Processing

TL;DR

This paper presents the maximal eigengap estimator, a novel method for accurately determining the direction of wideband acoustic signals using a single vector-sensor by optimally combining narrowband spectral data.

Contribution

It introduces the maximal eigengap estimator that optimally combines narrowband matrices to improve direction finding in wideband acoustic signals, demonstrating advantages over existing methods.

Findings

Effective in real-data acoustic applications

Maximizes signal-to-noise ratio across frequency band

Outperforms competing direction-of-arrival methods

Abstract

This paper introduces the maximal eigengap estimator for finding the direction of arrival of a wideband acoustic signal using a single vector-sensor. We show that in this setting narrowband cross-spectral density matrices can be combined in an optimal weighting that approximately maximizes signal-to-noise ratio across a wide frequency band. The signal subspace resulting from this optimal combination of narrowband power matrices defines the maximal eigengap estimator. We discuss the advantages of the maximal eigengap estimator over competing methods, and demonstrate its utility in a real-data application using signals collected in 2019 from an acoustic vector-sensor deployed in the Monterey Bay.