FIR filter design via spectral factorization and convex optimization (Q4952741)

The paper investigates the design of finite impulse response (FIR) filters satisfying upper and lower bounds on the frequency response magnitude. The associated optimization problems (called magnitude filter design problems), with the filter coefficients as variables and the frequency response bounds are constraints, are known to be non-convex in general. The purpose of the paper is to show that by a change of variables, a wide variety of magnitude FIR filter design problems can be rewritten, in terms of the autocorrelation coefficients, as (possibly nonlinear) convex semi-definite optimization problems. This transformation is obtained by using a change of variables: the new variables are the autocorrelation coefficients of the filter, recovered by spectral factorization subsequently. As a consequence, the original magnitude filter design problems can be (globally) solved with great efficiency on the basis of recently developed interior-point methods for convex optimization. The new method proves its tractability in solving a number of problems of practical interest such as filter and equalizer design, antenna array weight design, minimax decible problems. For the entire collection see [Zbl 0934.00020].