Character Recognition Experiments Using Unipen Data

This paper presents experiments that compare the performances of several versions of a Regional-Fuzzy Representation (RFR) developed for Cursive Handwriting Recognition (CHR). These experiments are conducted using a common Neural Network (NN) classifier, namely a MultiLayer Perceptron (MLP) trained with backpropagation. Results are given for Sections 1a (isolated digits), 1c (isolated lower-case), and part of Section 3 (lower-case extracted from phrases) of the Unipen database. Data set TrainR01/V07 is used for training while DevTest-R01/V02 is used for testing. The best overall representation yields recognition rates of respectively 97:0% and 85:6% for isolated digits and lower case, and 84:4% for lower-case extracted from phrases.