CHARACTER-RECOGNITION WITH NEURAL NETWORKS

Modeling neural networks is useful not only in understanding the mechanism of the brain, but also in obtaining new design principles for character recognition systems. With this approach, the author has proposed various models for visual pattern recognition. This paper introduces two of them: the 'neocognitron' and 'selective attention' models. The neocognitron is a hierarchical neural network model, capable of deformation-invariant pattern recognition. An alphanumeric character recognition system has been developed based on the neocognitron model. The selective attention model has not only forward but also backward connections in a hierarchical network. It has the ability to segment patterns, as well as the function of recognizing them. The principles of this selective attention model can be extended to be used for several applications: for example, the recognition and segmentation of connected characters in cursive handwriting of English words, and the recognition of Chinese characters.