Efficient Preprocessing of Strokes for Online Handwritten Gurmukhi Script

Abstract

In current era, a new technology called pen computing is emerging. It includes mobile devices and applications in which electronic pen with a writing pad is used as the main input tool. For implementing pen-computing applications, online handwriting recognition system should be used. Online handwriting recognition systems have been developed for various scripts around the world. Very few attempts have been made to build an online handwriting recognition system for Gurmukhi script. Pen-based input in an online handwriting recognition system allows people to write text and enter data in their own natural way of handwriting on an electronic pad.

This thesis is an attempt to develop a preprocessing model for strokes written in Gurmukhi script. The pen-based devices are not so common in India and one reason for that is the absence of localized applications. Developing an online handwriting recognition system for Gurmukhi scripts for such devices would play an important role in making these devices available and usable for the Indian society. Preprocessing will play an important role in increasing the accuracy of such online handwriting recognition system.

In this study, a model for preprocessing a Gurmukhi stroke is proposed. This model consists of 5 basic algorithms for preprocessing. These algorithms are Size Normalization and Centering of stroke, Interpolation of points in a stroke, Uniformizing of points in a stroke, Smoothing of a stroke and Resampling of points in a stroke. Prior to these algorithms, a basic step called Stroke Capturing is done, which samples data points along the trajectory of an input device (electronic pen or mouse) while the character is drawn.

Chapter 1 of this thesis gives brief introduction about the handwriting recognition systems. Online and offline handwriting recognition systems are then discussed in detail. Since current study is done for Gurmukhi script, a brief description of Gurmukhi characters, vowel modifiers and numbers is also given. One of the most important aspects while developing online handwriting recognition is the various issues that one can face during it. These issues are also given in detail in this chapter. In the end, a brief literature has been given in chronological order, for studying preprocessing techniques used in different online handwriting recognition system for many languages around the world.

Different algorithms have been proposed for the preprocessing activity in Chapter 2. Size normalization and centering of stroke (Algorithm 2.1) is required to standardize a stroke, i.e., putting a stroke in a particular size and in the center of the window. Interpolation (Algorithm 2.2) is required to add missing points in a stroke. Uniformizing (Algorithm 2.3) of points is done so that the adjacent points in a stroke are at a particular distance or greater than that. Smoothing (Algorithm 2.4) is done so that any wild point and sharp corners are removed from the stroke. Resampling of points (Algorithm 2.5) is done so that a stroke contains 64 equidistant points after preprocessing.

Chapter 3 of this thesis report contains a detailed discussion on all the problems that were faced while implementing the algorithms discussed in chapter 2 and snapshots of correction made to those problems. In the last section of this chapter, success rate of the preprocessing algorithms proposed in this thesis has been given and an overall success rate of 94.20 % has been achieved.

Conclusion and future scope related with this work has been given Chapter 4.