Arabic Handwritten Script Recognition System
Based on HOG and Gabor Features
Mohamed Elleuch1, Ansar Hani 2, and Monji
Kherallah3
1National School of Computer Science, University of
Manouba, Tunisia
2Faculty of Economics and Management of Sfax,
University of Sfax, Tunisia
3Faculty of Sciences, University of Sfax, Tunisia
Abstract: Considered as among the most thriving applications in the pattern
recognition field, handwriting recognition, despite being quite matured, it
still raises so many research questions which are a challenge for the Arabic
Handwritten Script. In this paper, we investigate Support Vector Machines (SVM)
for Arabic Handwritten Script recognition. The proposed method takes the
handcrafted feature as input and proceeds with a supervised learning algorithm.
As designed feature, Histogram of Oriented Gradients (HOG) is used to extract
feature vectors from textual images. The Multi-class SVM with an RBF kernel was
chosen and tested on Arabic Handwritten Database named IFN/ENIT. Performances
of the feature extraction method are compared with Gabor filter, showing the
effectiveness of the HOG descriptor. We present simulation results so that we
will be able to prove that the good functioning on the suggested system
based-SVM classifier.
Keywords: SVM, arabic handwritten recognition,
handcraft feature, IFN/ENIT, HOG.
Received February 10, 2017; accepted May 10, 2017