Comparative performance study of element-space and beam-space MUSIC estimators.pdf

CIRCUITS SYSTEMS SIGNAL PROCESS VOL. 10, NO. 3, 1991 COMPARATIVE PERFORMANCE STUDY OF ELEMENT-SPACE AND BEAM-SPACE M U S I C ESTIMATORS * Petre Stoica 1 and Arye Nehorai 2 Abstract. The goal of this paper is to show that in large samples the variances of the direction-of-arrival (DOA) estimates provided by the element-space MUSIC (i.e., MUSIC applied directly to the sensor output data) are always less than the corresponding variances associated with the beam-space MUSIC (i.e., MUSIC applied to spatially filtered sensor data). 1. Introduction Consider a wavefield generated by n na r row-band point sources and additive noise, impinging on a passive array of m sensors situated in the same plane with the sources. Let y(t)~ C ? 1 denote the sensor output vector at time instant t (t = t 1, t2 , . . . ) . A frequently used model of y(t) is the following [ I ]- [5]: y(t) = Ax(t) + e(t), (1.1) where e(t) ~ C ? 1 is an additive noise, x(t) is the complex envelope vector of the source signals, and A = [a(01) -.. a(0,)]. (1.2) In (1.2), 0 i is the direction-of-arrival ( D O A ) of the ith source signal, and a(Oi) ~ C ? a is the ar ray response vector to a unit wave impinging from direction 0i (a(O~) is also called the steering vector associated with the ith source ) . In D O A est imation problems, it is usually assumed that the a r ray is calibrated such that a(Oi) is a known function of the unknown 0~. Under this * Received March 27, 1990. The work ofA. Nehorai was supported by a grant from the Air Force Office of Scientific Research, under Grant No. AFOSR-90-0164 and the Office of Naval Research under Grant No. N00014-91-J-1298. 1 Department of Automatic Control, Polytechnic Institute of Bucharest, Splaiul Independen- tel 313, R-77 206 Bucharest, Romania. 2 Department of Electrical Engineering, Yale University, New Haven, Connecticut 06520, USA. 286 STOICA AND NEHORAI assumption of known array manifold, the eleme