HYBRID FUZZY-NEURAL SYSTEMS IN HANDWRITTEN WORD RECOGNITION

Two hybrid fuzzy neural systems are developed and applied to handwritt en word recognition. The word recognition system requires a module tha t assigns character class membership values to segments of images of h andwritten words, The module must accurately represent ambiguities bet ween character classes and assign low membership values to a wide vari ety of noncharacter segments resulting from erroneous segmentations, E ach hybrid is a cascaded system, The first stage of both is a self-org anizing feature map (SOFM), The second stages map distances into membe rship values, The third stage of one system is a multilayer perceptron (MLP), The third stage of the other is a bank of Choquet fuzzy integr als (FI's), The two systems are compared individually and as a combina tion to the baseline system, The new systems each perform better than the baseline system, The MLP system slightly outperforms the FI system ; but the combination of the two outperforms the individual systems wi th a small increase in computational cost over the MLP system, Recogni tion rates of over 92% are achieved with a lexicon set having average size of 100, Experiments were performed on a standard test set from th e SUNY/USPS CD-ROM database.