Pattern characteristics of an evolution between two classes

Abstract

Pattern recognition involves two phases: learning and recognition. Using fuzzy theory, the first phase consists mainly in learning membership functions of classes. The recognition phase (or diagnosis) consists in computing membership degrees of the data to classes and deciding in which class the data fits more appropriately. In real-time processes the system may evolve from one class to another. The main idea of this paper is to propose a method that is able to learn and recognize this evolution. The learning phase is based on fuzzy interval regressions. The second part presents a method that gives a diagnosis depending on the percentage of evolution. A decision method is suggested at the end.