Recognizing the isolated Arabic characters using neural networks

الملخص EN

This paper proposes a technique for recognizing isolated Arabic characters using neural networks.

This technique consists of three parts: body classifier, complementary classifier, and aggregate classifier.

The body classifier is designed to recognize the main body of the unknown character.

Hopfield network is used in this part to enhance the unknown character and to get rid of noise and associated complementary.

Furthermore, it uses a backpropagation network to recognize the main body of the enhanced unknown character.

The complementary classifier recognizes the number of dots or zigzag that are associated with the body of character and their position.

The aggregate classifier combines the results of the previous two classifiers and classifies the whole unknown character.

As a result of dividing the recognition process into three parts the number of patterns required to be recognized in each part has been reduced as well as speeds up the learning of neural network and increases the recognition rate.

The proposed technique has been implemented and shown a reasonable recognition rate.