OFFLINE HANDWRITTEN DEVANAGARI CHARACTER RECOGNITION USING FUSION OF CLASSIFIERS

Abstract

Now a days, there is need for the digitalization of handwritten documents. There is a large scope of research
in this area. Continuous improvement in Handwritten Character Recognition (HWCR) techniques milestones in this
research area. HWCR system is the software to accept and process the handwritten input images from sources such as
documents, photographs. HWCR is the ability to transform them to machine readable and editable format. Devanagari
script is used as base for various Indian languages such as Marathi, Sanskrit, Hindi, etc. and foreign languages such as
Nepali. The work proposed in our Handwritten Devanagari Characters Recognition System tries to automate recognition
of handwritten Devanagari isolated characters by ensembling different classifiers. Ensemble classifier is constructed by
using Support Vector Machine (SVM) [1], K-Nearest Neighbor (KNN) [1] and Neural Network (NN) [2] which increases
the performance by ensembling classifiers. The proposed system gives better results than individual classifiers.