https://jjcit.org/paper/32 ARABIC HAND-WRITTEN CHARACTER RECOGNITION BASED ON DEEP CONVOLUTIONAL NEURAL NETWORKS 10.5455/jjcit.71-1498142206 Khaled S. Younis Convolutional neural network,Deep learning,Optical character recognition,Arabic handwritten character recognition,EMNIST 54 1021 627 2017-06-24 2017-09-24 2017-10-31 The automatic analysis and recognition of offline Arabic handwritten characters from images is an important problem in many applications. Even with the great progress of recent research in optical character recognition, a few problems still wait to be solved, especially for Arabic characters. The emergence of Deep Neural Networks promises a strong solution to some of these problems. We present a deep neural network for the handwritten Arabic character recognition problem that uses convolutional neural network (CNN) models with regularization parameters such as batch normalization to prevent overfitting. We applied the Deep CNN for the AIA9k and the AHCD databases and the classification accuracies for the two datasets were 94.8% and 97.6%, respectively. A study of the network performance on the EMNIST and a form-based AHCD dataset were performed to aid in the analysis.