This paper presents a algorithm for identification of linear time-invariant nonminimum phase system from only the output data The input is independent identically distributed (i.i.d) non-Gaussian white noise and is unavailable. This approach is based on the idea of evaluating the bicepstrum of third-order cumulant of the observed output data. The minimum and maximum phase components are reconstructed separately. It is flexible enough to be applied on linear AR, MA and ARMA model without a priori knowledge of the type of the model. Simulation results verify the effectiveness of this algorithm for nonminimum phase system identification.
展开▼