Arabic character recognition using image density approach via error correcting neural network

There are many software techniques to recognize Arabic characters. On the other hand there are no hardware implementation, to recognize Arabic characters in different text positions, has been reported in the literature yet. Our goal in this paper is twofold (i) to develop a hardware implementation via neural networks for Arabic character recognition system. Image density to extract the character features is used to recognize different characters at different text positions. The backpropagation technique is applied for neural network learning. Error correcting neural network is used to achieve a performance accuracy of 100%. (ii) to compare this new approach with the sequential software implementation and the parallel software implementation with respect to accuracy, speed, and cost.