Cyclic Regression For Weighted Subspace Fitting To Find Multiple Signal Directions
James Lo, Neerchal K. Nagaraj, Andrew L. Rukhin · 2005
l An asymptotically equivalent expression of the weighted subspace fitting (WSF) criterion is obtained that does not involve signal or noise eigenvalues. Facilitated with the expression, a cyclic regression algorithm is used for finding the signal directions of arrival. It iterates a few linear regressions and data transformations. Simulation shows that the estimates generated by the algorithm are close to the ones optimal with respect to the original WSF criterion.