Asymptotic performance analysis of ESPRIT, higher order ESPRIT, and virtual ESPRIT algorithms

Abstract

In this paper, we present an asymptotic performance analysis of three subspace-based methods for direction of arrival (DOA) estimation-the ESPRIT algorithm using second order statistics, the higher order ESPRIT algorithm using fourth-order cumulants, and the virtual ESPRIT (VESPA) algorithm using fourth-order cumulants. We examine the least-squares version of these algorithms, derive the expressions for the asymptotic variance of the estimated DOAs, and use specific examples to compare the relative performance of the algorithms. Finally, we present Monte Carlo simulations to validate the theoretical analysis.