Arabic Handwritten Character Recognition Based on Convolution Neural Networks

Abstract

Automatic handwriting recognition is a useful task for many applications. The main Research has focused on the Latin languages. However, few approaches have been proposed for the Arabic language due to the specific and complex features of handwritten Arabic text. In this paper, we propose a Deep Learning (DL) approach for Arabic character recognition using proposed model of convolutional neural networks (CNN). In our work, we dealt with the specific features of Arabic text, in particular the variation of the shape of characters according to its position in the word based a new model of CNN network. In the experimental evaluation, we use hijja dataset in train and test steps. Obtained results prove the efficiency of our model, achieving accuracy of 95% on the Hijja dataset.