A new method for noise subspace estimation based on the spatial smoothing Lanczos algorithm

摘要

A new method is proposed to estimate a noise subspace. It is shown that the redundant prefilters of the multistage Wiener filter (MSWF) are capable of creating an orthogonal basis for the noise subspace. Based on the classical spatial smoothing technique and the Lanczos algorithm, a novel technique is presented to obtained the noise subspace in the case of coherent signals. The new estimator outperforms its counterparts in terms of computational complexity. Finally, the theoretical observations are illustrated by numerical results.