ANALYSIS OF FEATURE EXTRACTION AND CLASSIFICATION FOR OFFLINE ARABIC HANDWRITING WORD RECOGNITION

Arabic handwriting recognition is currently experiencing a massive rise in terms of analysis research. It is clear in the lecture review that researchers concentrate on Arabic handwriting recognition. A major increase has been reported in feature extraction and classification techniques compared to before. Offline Arabic handwriting recognition is an exceptionally significant research topic due to the cursive nature of written Arabic words and writing styles which makes the problem of recognizing Arabic words hard and challenging. In this paper, we investigate three feature extractors techniques Chebyshev Moments (CM), Statistical and Contour based Feature (SCF) and Zernike moment (ZER) for offline Arabic handwritten word recognition. We have considered a number of classes with different shapes of an image. These features have been trained separately with three classifiers namely, Extreme Learning Machine (ELM), Support Vector Machine (SVM) and Reduced Kernel Extreme Learning Machine (RKELM). The experiments were evaluated on IFN/ENIT database, experimental results show that shape of feature has a major impact on training classifiers.