Subspace tracking based on the projection approach and the recursive least squares method

The author presents a new algorithm for tracking the signal subspace recursively. It is based on a novel interpretation of the signal subspace as the solution of a projection like unconstrained minimization task. It is shown that the recursive least squares technique can be applied to solve this problem by approximation projections appropriately. The resulting algorithm has a computational complexity of O(nr) where n is the dimension of the problem and r is the number of desired eigencomponents, respectively. Simulation results show that the frequency tracking capability of this algorithm is virtually identical to and in some cases more robust than the more computationally expensive batch eigendecomposition.<>