A general class of recursive minimum variance distortionless response estimators

Abstract

In deterministic parameters estimation, it is common place to design a minimum variance distortionless response estimator (MVDRE) instead of a maximum likelihood estimator to tackle the problem of identifying the components of observations formed from a linear superposition of individual signals to noisy data. When several observations are available and the individual signals are allowed to perform a random walk between observations, one obtains the general class of linear discrete state-space models. This paper introduces a novel recursive formulation of the MVDREs of individual signals compatible with recursive estimation.