Arabic handwriting text recognition based on efficient segmentation, DCT and HOG features

Writing in its different forms, printed and manuscript has always been a tool essential in human communication, and as ubiquitous in most areas of its operations. It is used to store and archive knowledge. Thereby, human has always developed techniques for sustainability across generations. Indeed, with the advent of new information technologies and electronics computers, and further increase the power of machines, automated processing. Besides, one of the important system is the handwriting recognition. Handwriting text recognition system based on efficient segmentation, DCT and HOG Features is proposed. The proposed system depends on the segmentation of the text into words only. Moreover, the system achieved best recognition accuracy 96.317% based on the used methods and SVM classifier. © 2016 SERSC.