Decoupled Method for Approximation of Signals by Exponentials

Abstract

This noniterative method for approximating empirical signals over [O, ?) by a linear combination of exponentials yields suboptimal approximation. Notably, the dependence of the suboptimal exponents ?i? on the fractional integral square error ? is such that lim??0 ?i = ?i, the optimal exponents. The integral square error in representation is studied for a sin and a square pulse, and a useful error formula is developed.