Frequency-Domain Identification of Continuous-Time Output Error Models from Non-Uniformly Sampled Data

摘要

This paper treats direct identification of continuous-time autoregressive moving average (CARMA) time-series models. The main result is a method for estimating the continuous-time power spectral density from non-uniformly sampled data. It is based on the interpolation (smoothing) using the Kalman filter. A deeper analysis is also carried out for the case of uniformly sample ddata. This analysis provides a basis for proceeding with the non-uniform case. Numerical examples illustrating the performance of the method are also provided both, for spectral and subsequent parameter estimation.