Improving the resolution performance of eigenstructure-based direction-finding algorithms

Recently there has been much interest in algorithms which form a direction-finding spectrum based on the eigenstructure of the sensor covariance matrix. These algorithms are attractive because of their ability to achieve Cramer-Rao direction-finding accuracy bounds for closely spaced emitters, provided the available signal-to-noise (SNR) is high enough to resolve two distinct peaks in the estimated spectrum. In this paper we present several methods for reducing the SNR required for resolution. The first method involves examining the roots of the spectrum polynomial. This technique is applicable when a uniformly spaced sensor array is in use. The second method uses the properties of the so-called signal-space eigenvectors to define a rational (pole-zero) spectrum function with improved resolution capabilities.