Classification in Devanagari Script using Elliptical Region-wise Features

In this article, an attempt has been made to develop a system for classification of online handwritten text and non-text data from within a single online handwritten document in the most popular Indic script-Devanagari. As per our knowledge, no recognized work exists for handwritten text and non-text document classification in online mode in any Indic script. To develop this system an elliptical region-wise feature extraction approach has been proposed in this article. In this approach, each online stroke information of text and non-text documents is divided into smaller elliptical regions by constructing several concentric ellipses around the stroke. Each elliptical region is further divided into several sub-regions before extracting various structural and directional features of stroke portions from each sub region. These features are then studied in Hidden Markov Model (HMM) based classification platform. The efficiency of the present system has been measured on a self-generated dataset and it has provided promising result.