Tracking error bounds of adaptive nonstationary filtering (Q1059040): Difference between revisions

The paper deals with stochastic adaptive linear filtering algorithms. More precisely, its aim is to analyze the tracking capacity of such algorithms. It is assumed that the statistics of random signals \(a_ k\) and observation vectors \(x_ k\) are unknown. Moreover, the usual assumption of independent observations is replaced by a bounded memory model. Under these conditions it is shown that the mean square tracking deviation between the optimum vector and the algorithms output behaves in accordance with common intuition.