In this paper, a system is proposed that can recognize the handwritten numerals in dependent of its scale and position. The binary input image is transformed into a invariant image using the M transform function and neural network.
Training of the ne...
In this paper, a system is proposed that can recognize the handwritten numerals in dependent of its scale and position. The binary input image is transformed into a invariant image using the M transform function and neural network.
Training of the neural network is carried out by the descending epsilon method in the error back-propagation algorithm. Only 65 characters out of total 230 sample characters are used for training the neural network and the rest 165 characters for testing the generalization capability of the implemented system. As a result of this experiment, the recognition rate of handwritten Numbers was optimally achieved by 95.34% in case of using 65 characters for the training of the neural network.