Off-line cursive-script recognition using a neural network

Reports the applications of a feed-forward neural network classifier to a cursive-script recognition system. One of the difficulties of cursive-script recognition is the variability of styles of writers. Having this problem in mind and knowing the widely recognised ability of neural networks to generalise, the authors developed a cursive-script recognition system which takes a normalised geometrical features as inputs, and the outputs correspond to any letter of the English lower-case alphabet. The results for single and two writers are encouraging