Open AccessRobust DOA estimation of complex correlated signals in non-Gaussian CES distributed models
Author(s) -
Habti Abeida
Publication year - 2022
Publication title -
serbian journal of electrical engineering
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.133
H-Index - 5
eISSN - 2217-7183
pISSN - 1451-4869
DOI - 10.2298/sjee2201099a
Subject(s) - estimator , direction of arrival , robustness (evolution) , algorithm , covariance matrix , gaussian , covariance , subspace topology , mathematics , signal subspace , complex normal distribution , gaussian noise , computer science , noise (video) , statistics , artificial intelligence , physics , telecommunications , biochemistry , chemistry , quantum mechanics , antenna (radio) , image (mathematics) , gene
The sample covariance matrix (SCM) is commonly used in direction-of-arrival (DOA) estimation methods when the noise or observations are circular complex Gaussian (C CG) distributed. However, with a very heavy-tailed non-Gaussian noise model, the SCM-based DOA estimation methods fail to provide an accurate estimate of DOA. This paper presents a numerical analysis of the resolving capability of subspace-based circular (C) and non-circular (NC) multiple signal classification (MUSIC) DOA estimation methods of arbitrarily narrowband correlated signal sources corrupted by circular complex elliptical symmetric (C CES) distributed noise. It evaluates the robustness of these methods for correlated C and NC sources by employing the robust complex M-estimators instead of SCM. It study also the effects of correlation on robust MUSIC-based DOA estimation algorithms accuracy as a function of the magnitude and phase of the correlation coefficients. Simulations results show that the NC MUSIC algorithm which requires fewer sensor elements yields robust estimates of DOA for correlated sources than the C MUSIC algorithm using the M-estimators.