Recognition of Online Handwritten Gurmukhi Strokes using Support Vector Machine

Abstract

Communication is always an important part of our life, either in the form of speech or writing. Natural handwriting is one of the easiest ways to exchange information. In this world of technology exchanging information between user and computer is of immense importance and input devices such as keyboard and mouse have limitations as they cannot provide natural handwriting as input. Online handwriting recognition system can be used as an easiest and natural way of communication between user and human computers. Therefore pen-based interfaces are becoming more and more popular and hence a lot of research is being done for recognition. Research work presented in this thesis aims to recognize character with higher accuracy written in Gurmukhi script using Support Vector Machine (SVM) by improving processes of pre-processing phase used for recognition. Gurmukhi is a script of Punjabi Language which is widely used across the world. This thesis is divided into five chapters. A brief outline of each chapter is given in the following paragraph. First chapter of this report consists of introduction to online handwritten recognition system, issues in online handwritten recognition system overview of Gurmukhi script and literature review. Issues in online handwriting recognition system includes, handwriting style variations, constrained and unconstrained handwriting, personal, situational and material factors, writer dependent vs. writer independent recognition system. In the second chapter of literature review, a detailed literature survey on each phase of established procedure of online handwriting recognition has been presented. Third chapter gives the detailed work carried out in three phases. They are data collection, pre-processing and feature extraction. In data collection phase, input handwritten strokes are collected is shown. Phases of pre-processing are discussed. In the end feature extraction is explained. Fourth chapter describes the recognition techniques that can be used for online handwritten recognition system. In this work SVM is used as a classifier for the recognition of handwritten strokes. The results of the algorithm used in this thesis work are shown in the fifth Chapter. The cross-validation testing has been used for calculating the accuracy. I have worked on 2-fold, 3-fold 4-fold, 5-fold and 6-fold cross-validation testing on a sample of 30, 50 and 100 of each class. Finally, in the last chapter result of my work is represented. Future scope of the online work is also mentioned.