Handwritten Arabic words recognition using multi layer perceptron and Zernik moments
I. El-Fegh, Zakaria Suliman Zubi, Ali A. Elrowayati, Faraj A. El-Mouadib · 2009
This paper presents the application of Multi Layer Perceptron (MLP) Artificial Neural Network to classification of handwritten Arabic words. Zernik Moments are used as a feature vector for each word. An efficient way to select the most suitable order of Zernik moments is also presented. The MLP is trained in a supervised fashion using the Back Propagation learning algorithm. Having being trained, the MLP is tested on different set of handwritten Arabic words that has never been seen by the MLP. Several experiments are performed to select the best MLP structure. Experimental results have shown that with the presented structure and the order of the Zernik Moments more than 87% of correct recognition was obtained.