Knowledge-based understanding of signals: An introduction

Abstract

The paper provides an introductory review to the knowledge-based approach to signal understanding. Basic issues in knowledge-based system (KBS) design are presented following current methodologies, and peculiarities of interpretive problems are discussed in this context. Formal descriptions are given in terms of objects and tasks, the former denoting abstract data and concept representations, the latter generic knowledge-processing steps (numerical, inferential). Understanding patterns is seen as a sequence of task activations; control is a body of knowledge providing a system with dynamical focus of attention. Tasks are analysed by means of examples from the geophysical domain. Strong and weak points of the KBS approach are also discussed, to give an insight into achievements and current limitations: researchers in signal processing are expected to play an important rôle in future developments.