The Gauss-Markov Approximation Scheme for Identification of Multivariable Dynamic Systems Through Realization Theory: An Explicit Approach

摘要

How dynamic, linear, time invariant systems can be described in terms of the Hankel representation (H model). The H model and Markov parameters description of a system is then considered. Results of digital simulations are examined showing minimal variances in the Markov parameters as compared to a least squares estimation and an estimation using a sample covariance matrix. The use of the Ho-Kalman algorithm, based on a pseduo-inversion and singular valve decomposition method, to filter noisy Markov parameters is also explained.