Reweighted least squares design of fir filters

This paper presents a new application of the iterative reweighted least squared error algorithm to obtain an Lp approximation for designing a linear phase FIR filter. This method considers 1 ≤ p ≤ ∞ as being interesting itself or as a compromise between the traditional least squared error and Chebyshev designs. A new method is given which has good initial convergence coupled with quadratic final convergence. This performs much better than the Rice-Lawson algorithm. We also apply this method to designs with different p in different frequency bands and to solving the constrained least squared error problem.