In this paper, an Optical Character Recognition system for printed/scanned Pashto continuous text is presented.
The proposed Pashto optical character recognition system uses a Feed Forward Neural Network (FFNN), consisting
of an input layer, a hidden layer and an output layer. The input layer is composed of 315 neurons, which receive the
pixels data i.e. binary data from a 21x15 symbol pixel matrix. The hidden layer contains 2000 neurons which has
been chosen after testing based on optimal result, while the output layer is composed of 6 neurons. As the joinable
Pashto characters on different locations in text change its size and shapes, as a result 60 Pashto characters with
110 samples for each Pashto character has been used to train the network.