Novel Method for Handwriting Recognition Using a Variant of Decision Trees

Abstract

Recognizing the hand written document is one of the biggest challenges because of the variations in the handwritten text. The basic idea of this novel method is "divide and conquer". The algorithm recognizes the image in form of predefined glyphs. Based on the arrangement of the recognized glyphs, the exact character in the image can be recognized by comparing with the predefined glyphs using pattern trees. These set of predefined pattern trees is the signature of each character. The input symbol is compared with each of the patterns and the highest match is found and the correct character is identified.