Spectral estimator with rank-deficient covariance matrix

A spectral estimator based on the rank-deficient sample covariance matrix was developed to improve the robustness of estimates of the rank-deficient robust Capon filter-bank (RCF) spectral estimator. The estimator utilizes constraints on both the subspace and the errors of steering vectors to design a narrow filter. The filter outputs the spectral estimate with the filtering vector as its input. The filtering vector is formed by weighting snapshots with the Capon weight. Simulation results for a singular sample covariance matrix show that the spectral estimator obtains the spectral estimates with better accuracy and resolution than the rank-deficient RCF spectral estimator. Thus the estimator is an efficient estimator which utilizes only a few snapshots to estimate the spectrum.