Using fuzzy filters as feature detectors

A neuro-fuzzy model of adaptive learning and feature detection is presented. The model, called the fuzzy filtered neural network, was first introduced in a previous publication, which showed its validity in the domain of plasma analysis. Here the authors extend the model to another problem, the recognition of hand-written numerals, to demonstrate its generality. The authors propose three versions of the architecture, which use one-dimensional fuzzy filters, two-dimensional fuzzy filters, and genetic-algorithm-based fuzzy filters, respectively, as feature detectors. All three versions smoothly handle such issues of a real-world pattern recognition problem as drifting and noise. Simulation results show that the proposed model is an efficient architecture for achieving high recognition accuracy