Distinguish of Arabic symbols printed using Cohen's neural network based on the histogram

In this research histogram technique used lo extract the characteristic for. the images of Arabic characters, and evaluate their histogram values which used with Kohonen neural network as input, for recognizing the Arabic characters using intelligence methods.<br /> The Kohonen neural network has been trained on vertical histogramvalues for a set of letters including (ث ،ح ،د،ر،س ،ض ،ط ،غ ،ق ،ن) printed with Simplified Arabic font type, size 14, and then it has been test the network on another set of character that has not been trained by the network, the results were converging very well, and also it has been test on a set of noisy letters images, where the results were achieved true 100% , and then it has been test on a set of larger size character images (size 18 point), and also the results were not different with the previous. Also, it has been trained the network by using horizontal histogram for the same group of characters, and test the network on another set of characters that has not been trained by the network, and the results had lower closeness from the results that obtained with vertical histogram, and the closeness was more decreasing when has been testing on a set of noisy letters images, and also it has been test on a set of larger size letters images (size 18 point), and the results were achieved true 66% .<br /> So, it is discern previously that depending on vertical histogram values gives higher closeness from horizontal histogram vales. And when.<br /> depending on the two histograms together, it was gives bad results with other neural network as well as kohonen neural network.