Broadband Sparse Array Focusing Via Spatial Periodogram Averaging and Correlation Resampling

TL;DR

This paper introduces two broadband focusing algorithms for sparse array data that enhance source detection and direction finding by averaging spatial periodograms or resampling correlations across frequency bands.

Contribution

The paper presents novel coherent broadband focusing algorithms using spatial periodogram averaging and correlation resampling for sparse arrays, improving source enumeration and DOA estimation.

Findings

Both algorithms outperform incoherent focusing in limited snapshot scenarios.

The methods improve source enumeration accuracy.

Enhanced DOA estimation performance demonstrated through simulations.

Abstract

This paper proposes two coherent broadband focusing algorithms for spatial correlation estimation using sparse linear arrays. Both algorithms decompose the time-domain array data into disjoint frequency bands through discrete Fourier transform or filter banks to obtain broadband frequency-domain snapshots. The periodogram averaging (AP) algorithm starts in the frequency domain by estimating the broadband spatial periodograms for all bands and then averaging them to reinforce the sources' spatial spectral information. Taking inverse spatial Fourier transform of the combined spatial periodogram estimates the focused spatial correlations. Alternatively, the spatial correlation resampling (SCR) algorithm directly computes the spatial correlations for each band and then rescales the spatial sampling rate to align at a focused frequency. The resampled spatial correlations from all frequency bands are then averaged to estimate the focused spatial correlations. The spatial correlations estimated from the AP or SCR algorithms populate the diagonals of a Hermitian Toeplitz augmented covariance matrix (ACM). The focused ACM is the input of a new minimum description length (MDL) based criteria, termed MDL-gap, for source enumeration and the standard narrowband MUSIC algorithm for DOA estimation. Numerical simulations show that both the AP and SCR algorithms improve source enumeration and DOA estimation performances over the incoherent subspace focusing algorithm in snapshot limited scenarios.