Construction of a feature set for character recognition

A character recognition system which utilized neural network was trained to recognize handwritten alphanumeric characters. Unsupervised learning was adopted in order to ensure the necessary features for recognition were determined by the network in accordance with its structure. An algorithm which forced the network to pick up features with simple shapes was introduced so that the feature extraction mechanism became similar to that of the visual system. The features determined through such process resulted in a set of texton patterns. The recognition ability of the network was so sufficient that a possibility of character recognition based on texton patterns is suggested.