Effective Recognition of Handwritten Arabic Text

Abstract

This paper presents an effective method for segmenting handwritten Arabic words based on projection histograms, as well as a comprehensive study of overlap letters, which represent a major problem in the recognition of Arabic script. The characters resulting from segmentation are recognized by an SVM classifier that was trained on a corpus of handwritten Arabic characters that also includes the main overlap letters identified in Arabic script. The proposed method was tested using the IFN/ENIT database and encouraging results were obtained.