Direction finding using beamspace root estimator banks

Motivated by the superior performance and reduced computational complexity of beamspace and root implementations of eigenstructure techniques, a beamspace root modification of the pseudorandom joint estimation strategy (PR-JES) is developed. The essence of the PR-JES is to generate the eigenstructure-based estimator bank for a given sample covariance or data matrix. Appropriately combining the results of "parallel" underlying estimators, the PR-JES removes outliers and improves the threshold performance. In the case of a nonuniform array, the interpolated array approach is exploited to enable the application of root underlying techniques. Simulations show that the proposed beamspace root implementation outperforms spectral elementspace PR-JES significantly and achieves a performance similar to or better than that of the stochastic maximum likelihood (ML) method.