A framework for knowledge based spectrum analysis

Abstract

A general framework for the development of a knowledge-based spectrum analysis system is presented. The knowledge available for rating different methods is limited by the trend in literature of comparing a new spectrum analysis method with the best of the competing methods. The authors have offset this problem by comparing all the methods within a class of spectrum estimators on the same performance index. The use of the repertory grid allows the spectrum analysis knowledge to be validated and can be used to limit the search space when a spectrum analysis technique fails by utilizing the similarities among different methods. The additive clustering model, being nonhierarchical, is shown to be well suited for establishing the similarities among different spectrum analysis methods within a specific class. An extension of the method to operate directly on the ratings in the repertory grid is provided. Strategy is used as a basis of the development of the inference engine in this work. In general, the use of strategy improves the quality of the designed knowledge-based system by providing an environment in which a limitless number of methods and data validation methods can be incorporated. Considering that the designed knowledge-based system can be used to study the spectrum analysis strategy, such a system can be used to conduct spectrum analysis methodological research. >