Automatic Cheque Number Recognition using Neural Network
K. Bramara Neelima, Thangavelu Saravanan · 2018
This work is aimed to design an intelligent cheque number recognition with an artificial neural network for automatic Cheque processing. For automatic cheque number recognition, a back propagation artificial neural network with single input layer, one hidden layer and single output layer is implemented. The work aimed to analyze the performance of the system by varying the regularization parameter i.e learning rate. Performance of the system is also analyzed using the learning curves to determine whether it has high bias or high variance. From the result of learning curve the optimal training sets required for the accurate recognition is decided. The performance of the system is analyzed by using correct detection which is defined as the ratio of correctly recognized digits to the total number of digits in the test set. The computational complexity for the training of 2094 data set is achieved around 909 seconds.