ABSTRACT
This chapter begins the "off-line" form which originally appeared in. Understanding this off-line form is essential towards developing its "on-line" counterpart, which takes the form of an adaptive filtering algorithm. The chapter traces the family of algorithms to an idea of Steiglitz and McBride for identifying the parameters of a linear system. It deduces the set of transfer functions which correspond to the stationary points of the Steiglitz-McBride iteration. The chapter suggests that the stationary points are still characterized by Theorem. It shows that in the sufficient-order case, the algorithm converges to the correct system identification if the disturbance term is white noise. The chapter assumes that for arbitrarily poor signal-to-noise ratios, the spectral characteristics of the disturbance term can always be chosen such that no stationary point exists. It suggests that the set of stationary points can be deduced by way of an interpolation problem.
