Identification of errors-in-variables models based on frequency domain power spectra

Identification of errors-in-variables (EIV) models is studied under the assumption that a discrete frequency domain power spectrum is given. A problem setting for identifying dynamic EIV models is presented, assuming that an output error is bounded and that the input noise level is low. An identification algorithm for the problem is given via a subspace identification method and J-spectral factorization techniques, where an associated Popov function and Riccati equation are derived. Numerical simulation results are also shown