Joint 2D DOA and spectrum estimation of incoherently distributed wideband sources via frequency-domain matching

Abstract

This paper aims at the problem of joint two-dimensional (2D) direction-of-arrival (DOA) and frequency-spectrum estimation of incoherently distributed wideband (IDW) sources. The typical methods of 2D DOA estimation for distributed sources are based on the narrowband-source assumption and usually require multiple eigenvalue decomposition, followed by exhaustive searching-based pair-matching, which is computationally complicated. In this paper, we extend the existing scheme to wideband sources and propose a simpler pair-matching method based on the frequency-domain parameters including central frequency and frequency spectrum. First, the 2D IDW source estimation problem is decomposed into two one-dimensional (1D) estimation problems by dividing an L-shape array into two orthogonal uniform linear arrays (ULAs) and then estimating the joint angular-frequency distribution (JAFD) matrices through the low-rank recovery method along the two ULAs. We further proposed a pairing scheme to match the individual source estimates from different JAFD matrices based on the fact that the estimated frequency spectrum on the same source is identical. The proposed method achieves a 100% accuracy rate of successful pairing in the 100 Monte-Carlo experiments when signal-to-noise ratio SNR=5 dB, which is higher than the central frequency paired method at least 10%. The root-mean-square-error (RMSE) results based on parametric and real-world source signals also demonstrate the effectiveness of the proposed method comparing with the traditional subspace methods and the central frequency paired method.