Hand Written Digits Recognition Using Digital Learning Networks

A pattern recognition system based on the n-tuple technique is developed and evaluated for use in classifying non-deterministic data with particular reference to unconstrained hand-written numerals. The system presented in this research fulfills the requirements of simplicity and efficiency making it attractive to practical use in present day industrial environments . This simplicity of operation is afford by the self evolving nature of the classifier since it is based on a training phase where the recognition logic is developed. Results are reported on the task of recognizing handwritten digits without any advanced pre-processing. The result are obtained by using a RAM-based neural network, making use of small receptive fields. Furthermore, a technique that introduces weights into the RAM net is reported.