Fuzzy methods for classification in diagnostic systems

Abstract

Computer-assisted interpretation of sensor data in a modern diagnostic system brings together various technologies, including signal and image processing, pattern recognition, and artificial intelligence. In both biomedical and advanced manufacturing applications the complexity of the processes under investigation renders the task of automated classification difficult. A method for reasoning in the presence of uncertainty in essential to accomplishing this task effectively. Recently the theory of fuzzy sets and related uncertainty calculi have emerged as viable alternatives to traditional probabilistic approaches in some applications. This paper is a survey of methods employing some concept of fuzziness or degree of membership in the solution of pattern classification problems, and the potential role of these methods in diagnostic systems.