scientific article; zbMATH DE number 1058108

From MaRDI portal
Publication:4353221





The paper gives an overview of frequency domain least squares estimators for rational transfer models of linear time-invariant systems. The statistical performance of different approaches are analyzed through their equivalent cost functions which take their minimum for the total least squares (TLS) method. By analyzing the cost function the authors show the statistical properties of the TLS estimator. They propose weighted TLS versions with nearly maximum likelihood (ML) properties which are obtained by comparison of TLS techniques with the ML approach. The main ideas in the development and analysis of TLS estimators are explained in section 2 (single input, single output systems). Section 3 described a generalization to the multi variable case. Section 4 studies the properties of the generalized total least square (GTLS) estimators when the true noise covariance matrix is replaced by the noise sample covariance matrix. The theory is illustrated in section 5 and 6 by simulation and real measurement examples. For the entire collection see [Zbl 0861.00018].











This page was built for publication:

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4353221)