Character segmentation in Malayalam Handwritten documents

Abstract

This paper propose a method to segment individual characters from a Malayalam Handwritten document. All the existing systems for Malayalam handwritten character recognition segments isolated characters from a document and deals only with a subset of characters. The proposed method takes the words from the original handwritten document for segmentation. So, here the method has to deal with touching characters, breaks in the characters, different handwriting styles, fonts etc. Character segmentation is done by using the combination of Vertical Projection method and connected component analysis method. Then the unique features for identifying each characters are extracted and given to a classifier. The classifier identifies each character and corresponding Malayalam character is given as output.