Subspace identification methods: review and re-interpretation

The paper is devoted for a systematic discussion on the subspace based identification methods. The structure of the noise components acting additively at various system signals is considered as the guiding principle for the analysis of the estimation algorithms. It is shown that replacing the state space approach typically applied in this field by a discrete time transfer function system realization the highly involved machinery of the subspace algorithms can be handled in a clearly ordered framework exhibited by a set of over-determined equations to be solved.