|Timothy A. Barton and Steven T. Smith
MIT Lincoln Laboratory
244 Wood Street
Lexington, MA 02173-9108
tel: (781) 981-3278
Abstract Adaptive nulling algorithms require a good estimate of the interference covariance matrix. In situations with limited sample support, such an estimate is not available unless there is covariance structure to be exploited. In applications such as space-time adaptive processing (STAP), where one may be attempting to null clutter in range and doppler, the underlying covariance matrix may be structured (e.g., block Toeplitz), and it is possible to exploit this structure to arrive at improved covariance estimates. Several structured covariance matrix estimators have been proposed for this purpose, including estimators that project the sample covariance matrix into the structured covariance matrix constraint set, as well as maximum likelihood (ML) estimators. The efficacy of several of these are analyzed in this talk in the context of a fully optimum STAP algorithm as well as several reduced-dimension sub-optimum STAP algorithms. The SINR losses resulting from these different algorithms are compared. In particular, an example illustrating the superior performance resulting from a new maximum likelihood algorithm (based on the expectation-maximization (EM) algorithm) is demonstrated using simulation and experimental data from the Mountaintop data set. It is shown specifically that for a low sample support, positive definite, sample covariance matrix, the SINR losses achieved away from main lobe clutter can be much greater for the projection covariance matrix estimators than that achieved using the EM algorithm-based ML estimator. In addition, the doppler coverage near main lobe clutter may also be significantly reduced when using the considered non-ML covariance matrix estimators.
Direct comments and questions to: email@example.com
© MIT Lincoln Laboratory. All rights reserved.