Plus d'informations ici
The Parks McClellan algorithm: a scalable approach for designing FIR filters
Intervenant(s) : Silviu Filip (ENS de Lyon) Based on the theory of minimax polynomial approximation, the Parks-McClellan algorithm is one of the most well-known approaches for designing finite impulse response filters. In this talk, I will give new insights on the practical behavior of this algorithm. The main ideas I will be focusing on are barycentric interpolation methods and numerically stable root finding routines that rely on determining the eigenvalues of appropriate structured matrices. Barycentric interpolation and eigenvalue-based rootfinders have garnered a lot of interest from numerical analysts in the last fifteen years. I will explain how these building blocks can be used to write an efficient implementation of the aforementioned algorithm, and showcase its robustness by comparing it to other filter design routines found inside widely used software environments, like Matlab and Scipy.
marc (at) null