An iterative subspace estimation method for adaptive array processing

The author presents a novel, low-complexity, iterative method of vector subspace estimation applicable to adaptive beamforming and spatial spectral estimation. This method combines two complementary procedures: the power/deflation method of principal eigenvector estimation and the Lanczos method of tridiagonalization for computing the orthonormal basis of a Krylov subspace. Side information concerning the spatial coherence of the deflated array data is developed to aid discrimination between the (coherent) signal and (noncoherent) noise subspaces. The effectiveness of the proposed method is demonstrated by processing a typical example of underwater acoustic receiver data based on N.L. Owsley's (1991) Standard Test Case. The algorithm's performance is compared with a baseline method that uses full singular value decomposition of the data.<>