Hand-printed character recognition system using artificial neural networks

A new technique is proposed for the recognition of hand-printed Latin characters using artificial neural networks together with conventional techniques. One advantage of this technique is that it combines rule-based (structural) and classification tests. Another advantage is that it it more efficient for large and complex sets. The technique can be divided into five major steps: (1) digitization of the image; (2) thinning of the binary image using a parallel thinning algorithm; tracing of the tree; (3) tracing of the skeleton of the image and construction of a binary tree; (4) extraction of features from the structural information; and (5) classification of the segmented descriptions as particular characters by a feedforward neural network trained by backpropagation