Frequency domain identification of MIMO state space models using the EM algorithm

In this paper we consider the problem of estimating a state-space model for Multiple-Input Multiple-Output systems. We show how the Expectation-Maximisation algorithm can be used in the frequency domain to obtain a Maximum Likelihood estimate for state-space models. This method can be applied to obtain robust estimates of the model focusing on a specific frequency band of the input/output data spectra. The proposed method can be used to identify systems excited with general input signals and it is able to deal with transients responses.