Filter diagonalization: Filtering and postprocessing with prolates

Abstract

A detailed account is given of a recent modification of the Filter Diagonalization technique that serves to analyze a signal spectrum within a selected energy range. Our approach employs for filtering the eigenfunctions of the Finite Fourier Transform, or prolates, which are superior to other filters due to their special properties. In particular, prolates are simultaneously band-limited and highly concentrated at a finite time-interval, producing filters with optimal accuracy. In addition both features are acquired by the convolution of a band-limited function with a prolate, that permits the latter to be interpolated via the Walter and Shen sampling formula, which essentially simplifies the supplementary computations. Rigorous filtering error estimates are obtained. Test calculations illustrate the facilities of the presented modification.