Handwritten Devanagari Numeral Recognition using Structural and Statistical Features 1

Abstract: In this paper, the methodology for off-line isolated handwritten Devanagari numeral recognition is proposed. The proposed methodology is based on the structural and statistical feature extraction techniques. Numeral recognition is the important field in image processing and pattern recognition. Handwritten Numeral recognition has received extensive attention in academic and engineering fields. In this proposed method of Hindi Numeral recognition the pre-processing techniques except Binarization like thinning, slant removal are minimized. In this work structural features on the Scanned images of isolated Hindi numerals are applied like left open, Right open, above open,Vertical Crossing, horizontal crossing and so on. These features will help in primary classification of Hindi numeral set and after primary classification more features to recognize individual Hindi Numeral are applied and get the accuracy 96.80%