Reference: Drakopoulos, J. A. & Hayes-Roth, B. tFPR: A Fuzzy and Structural Pattern Recognition System of Multi-Variate Time Dependent Patterns based on Sigmoidal Functions. Knowledge Systems Laboratory, May, 1994.
Abstract: tFPR is a fuzzy pattern recognition system that utilizes sigmoidals to approximate general truth evaluation functions (or fuzzy set graded membership functions) and can deal with time or other parameter-space-dependent patterns. Besides the use of sigmoidals to model membership functions a novel feature introduced in tFPR is its separation of pattern specification and evaluation processes. Given a set of input data and a pattern specification, tFPR evaluates the pattern for the input data by producing a continuous-valued output representing the possibility that the pattern appears in the current input data. The input data may be a number of time-dependent signals whose past values may influence the evaluation of the pattern. As a result, the system must deal with curves rather than points in the input domain. tFPR has been implemented in the BB1 blackboard architecture. It is being applied in a system for medical monitoring and will be used in similar monitoring systems for semiconductor fabrication and equipment monitoring.
Notes: Updated September 1994.
Full paper available as hqx.