Publication

Abstract : Handwritten Roman characters and numbers have been intensively examined in the past several decades, with satisfactory results. The Devanagari script, however, does not fit this description. One of the most often used scripts in India is Devanagari. This research work proposes designing of an effective Devanagari script handwriting digit recognition algorithm. The approach for off-line isolated handwritten Devanagari numeral recognition is put forward in this paper. The suggested methodology is based on techniques for extracting structural and statistical features. The crucial area of image processing and pattern recognition is numerical recognition. The disciplines of engineering and academia have given much attention to handwritten numeral recognition. Attributes like right open, above open, left open, vertical crossing, These attributes will aid in the initial categorization of the Hindi number set. This research work is to implement Convolution Neural Network (CNN) and Multi Layer Perceptron (MLP) for the initial classification, additional features are used to recognize each particular numerals. Project shows 98% accuracy and identification of area under ROC curve.