Wideband Modal Orthogonality: A New Approach for Broadband DOA Estimation

TL;DR

This paper introduces WIMO, a low-complexity broadband DOA estimation method that does not require prior source information, offering improved resolution, accuracy, and computational efficiency over existing techniques.

Contribution

The paper proposes the WIMO framework, a novel subspace-based approach for broadband DOA estimation that avoids spectral decomposition and prior source information.

Findings

WIMO achieves higher resolution and lower estimation error.

WIMO significantly reduces computational runtime.

WIMO outperforms state-of-the-art methods in numerical tests.

Abstract

Wideband direction of arrival (DOA) estimation techniques for sensors array have been studied extensively in the literature. Nevertheless, needing prior information on the number and directions of sources or demanding heavy computational load makes most of these techniques less useful in practice. In this paper, a low complexity subspace-based framework for DOA estimation of broadband signals, named as wideband modal orthogonality (WIMO), is proposed and accordingly two DOA estimators are developed. First, a closed-form approximation of spatial-temporal covariance matrix (STCM) in the uniform spectrum case is presented. The eigenvectors of STCM associated with non-zero eigenvalues are modal components of the wideband source in a given bandwidth and direction. WIMO idea is to extract these eigenvectors at desired DOAs from the approximated STCM and test their orthogonality to estimated noise subspace. In the non-uniform spectrum case, WIMO idea can be applied by approximating STCM through numerical integration. Fortunately, STCM approximation and modal extraction can be performed offline. WIMO provides DOA estimation without the conventional prerequisites, such as spectral decomposition, focusing procedure and, a priori information on the number of sources and their DOAs. Several numerical examples are conducted to compare the WIMO performance with the state-of-the-art methods. Simulations demonstrate that the two proposed DOA estimators achieve superior performance in terms of probability of resolution and estimation error along with orders of magnitude runtime speedup.