Subspace Estimation Without Eigenvectors for Adaptive Beamforming

摘要

This report describes a new fast projection technique for adaptive beamforming.It is based on a technique proposed by Yeh where Gram-Schmidt orthogonalization is applied to N rows of the covariance matrix to obtain the signal subspace. In Yeh's technique, N is equal to L the number of jammers. This technique, applied to bearing estimation, requires a priori knowledge of the number of sources. Even though it gives good results when the number of sources is much smaller than the number of array elements, the performance of Yeh's technique degrades severely as the number of jammers increases. We improve the technique by taking a larger number of rows in the orthogonalization: we add a thresholding procedure to the technique in order to optimize N. We compared this new technique, called Subspace Estimation without Eigenvectors (SEWE), with other fast projection techniques. SEWE was shown to give superior performance in terms of the Signal-to-Noise-plus-Jammer Ratio achievable with a given computational load, except when a very high SNJR is required.