Pattern and Character Recognition Systems-Picture Processing By Nets of Neuron-Like Elements

Spatial pattern recognition, of which the recognition of alpha numeric characters is a subclass, is an important and practical problem. More efficient coding of transmitted pictorial information and more efficient utilization of humanly produced information could result from its solution. The problem of pattern recognition has been stated as the assignment of a meaningful code to a recognizable structure in a set of signals. The signals, in this case organized spatially, are the result of a transformation from a visual picture field P to an electrical representation of this field. The points of this signal field S correspond to a characteristic of points in the picture. In this study, the reflectivity of given picture-point areas is quantized as black or white as a basis for a two-state electrical-signal representation of points of a pattern.