In this paper, we propose a method to improve recognition rate for off-line recognition of hand written numerals using a simple neural network trained with backpropagation algorithm. We use 4 direction features using 8 direction chain code and ring-pr...
In this paper, we propose a method to improve recognition rate for off-line recognition of hand written numerals using a simple neural network trained with backpropagation algorithm. We use 4 direction features using 8 direction chain code and ring-projection feature, fan mesh features using fing. Reliability factor is introduced reduce error rate. In order to verify the performance of the proposed method. exprements with handwritten numeral database of Concordia University of Canada were performed. This method is shown 95.2% of Recognition rate When used Reliability factor.