Abstract
A new Krylov subspace projection method, called intelligent multipoint Arnoldi (IMA) method, to approximate FIR filters by low-order linear-phase IIR filters is developed. Based on special realizations of FIR filters, iterative nature of Arnoldi algorithm, and expansion points selection scheme, the proposed method is very efficient in terms of computational complexity. In addition, the lower-order IIR filter can truly reflect dynamical features of the FIR filter and indeed satisfies original design specifications. Fundamental developments of the proposed IMA algorithm will be discussed in details. Numerical examples will demonstrate the accuracy and the efficiency of the proposed method.