A framework for the evaluation of spectral analysis techniques
Manuel Duarte Ortigueira, Jose M. Tribolet · 2005
The area of Spectral Analysis has grown considerably in recent years. A multitude of methods are now available and yet no tools have been developed to guide the naïve user in the difficult process of choosing an adequate technique for his particular data problem. The purpose of this paper is to address, within a very general framework, the problem of choosing the "best" Spectral Analysis technique, from within a set of available methods and for a given class of data signals, according to some pre- -specified criteria. The foundations of the evaluation framework we propose are laid out in this paper. An in-depth study based on a particular evaluation methodology is reported, using a goodness of fit criterion derived from the works of Kedem and Slud [4,5]. Examples of this approach will be given for several data cases and several spectral analysis methods.