On the advantages of subspace tracking for temporal updating

Abstract

When doing parameter estimation, ESPRIT is an often used algorithm. As input, ESPRIT needs the eigenvectors of the signal subspace. These can be generated by an eigenvalue decomposition or by subspace tracking algorithms. In this paper, we demonstrate, that subspace tracking is superior not only regarding computational complexity but also regarding accuracy, which is due to its temporal updating character. We will also detail the circumstances for which this holds and that delay is a side-effect.