Recognition of online and offline handwritten composite characters and numerals in Devnagari script

Abstract

The offline and online recognition of Devnagari handwritten basic, composite and numeral characters is performed. Holistic approach is proposed. Different statistical features like Zoning, Directional Distance newlineDistribution, Zernike moments, Discrete Cosine Transform, Gabor filters, Gradient and Direction Chaincode are implemented and applied with SVM classifier. Integrated feature approach is proposed. Both stroke-based and character-based methodology is newlineperformed for online recognition. A database of 224826 offline characters with 707 classes and 7200 online characters with 110 characters and 10 numeral classes is created for testing. Standard benchmark database is used for offline numeral recognition. newline