Structured total least norm and approximate GCDs of inexact polynomials

The determination of an approximate greatest common divisor (GCD) of two inexact polynomials f=f(y)$f=f(y)$ and g=g(y)$g=g(y)$ arises in several applications, including signal processing and control. This approximate GCD can be obtained by computing a structured low rank approximation S^*(f,g)$S^{*}(f,g)$ of the Sylvester resultant matrix S(f,g)$S(f,g)$. In this paper, the method of structured total least norm (STLN) is used to compute a low rank approximation of S(f,g)$S(f,g)$, and it is shown that important issues that have a considerable effect on the approximate GCD have not been considered. For example, the established works only yield one matrix S^*(f,g)$S^{*}(f,g)$, and therefore one approximate GCD, but it is shown in this paper that a family of structured low rank approximations can be computed, each member of which yields a different approximate GCD. Examples that illustrate the importance of these and other issues are presented.