On time-invariant realizations of discrete random processes

A class of discrete second-order random processes, described by linear time-invariant state-space models, is investigated. Equivalent realizations with reduced number of noise inputs are presented. In contrast to the innovations approach these realizations have time-invariant parameters.