Title: A novel Arabic font recognition system based on texture feature and dynamic training

Authors: Faten Kallel Jaiem; Monji Kherallah

Addresses: Research Groups on Intelligent Machines Lab, University of Sfax, Sfax, 3003, Tunisia ' Research Groups on Intelligent Machines Lab, University of Sfax, Sfax, 3003, Tunisia

Abstract: Recognising an Arabic text with OCR is a complex task caused by the cursive nature of Arabic script for printed and handwritten text. The Arabic letters change forms according to not only their position in the word, but also their font. In fact, developing a font recognition system as a pre-recognition step may help to increase the OCR performances. In this paper, we present an Arabic font recognition system using curvelet transform for feature extraction. Moreover, we expose a new classification strategy based on a back-propagation artificial neural network (BpANN) called a dynamics multi-BpANN-1Class classifier. To validate our proposed system, we first focused our research on a comparative study of five texture analysis techniques. Second, we compared our classifier to a classical BpANN. And finally, we validate the dynamic training for the classification phase.

Keywords: grey level cooccurrence matrix; GLCM; Gabor filter; wavelet; steerable pyramid; curvelet transform; Arabic font recognition; texture analysis techniques; dynamics multi-BpANN-1Class; BpANN.

DOI: 10.1504/IJISTA.2017.088053

International Journal of Intelligent Systems Technologies and Applications, 2017 Vol.16 No.4, pp.289 - 308

Available online: 17 Nov 2017 *

Full-text access for editors Access for subscribers Purchase this article Comment on this article