Recognition of handwritten special characters and Gurmukhi numerals using artificial neural networks

Abstract

Online handwriting recognition process is a quick and natural way of communication between computer and human beings. Handwritten character recognition has been one of the most important and challenging research areas in the field of pattern recognition. Handwritten character recognition (HCR) is the process of converting handwritten text into machine processable format. It contributes immensely to the advancement of an automation process and can improve the interface between human and machine in various applications. Variations in handwriting make it difficult and challenging to achieve a high degree of accuracy. This dissertation aims to develop an online system for the recognition of handwritten Gurmukhi numerals and special characters. Gurmukhi is the script of Punjabi language which is widely spoken across the globe. For the recognition of characters/numerals, neural networks have been applied. The study focuses on the recognition of handwritten and printed characters; and the results emerging from it are presented. Handwritten character recognition has been a popular research area for many years because of its various applications. These applications are quite useful for the organizations such as railway, embassies, etc. where both English and regional languages are used. Many forms and applications are filled in regional languages. Sometimes these forms are scanned directly. HCR is a process of automatic computer recognition of characters in optically scanned and digitized pages of text. It can be online or offline. In online numeral and special character recognition, data are captured during the writing process with the help of a special pen and an electronic interface. Offline documents are scanned images of pre-written text, generally, on sheet or paper. Offline number recognition is significantly different from online recognition, because here stoke information is not available. For Gurmukhi numerals recognition, the accuracy achieved during training mode is 96.4%; and for testing mode it is 89.6%. For special characters recognition, training mode accuracy is 97.1%; while it is 92.0% for testing mode. The present study has been divided into six chapters. A brief review of these chapters is given below. Chapter-1 presents the different concepts of handwritten character recognition and also those of artificial neural networks. Chapter-2 examines and evaluates the existing literature on the subject. The different neural network techniques used in HCR have been reviewed to identify the gap in research area. Chapter-3 demonstrates how we can use the ANNs for approximating mathematical functions. Chapter-4 explains the methods used for data collection and details on online handwritten Gurmukhi numeral recognition using ANN. Chapter-5 provides the description of data collection for online handwritten special character recognition using ANN. Chapter-6 concludes the work by developing a system for handwritten character/numeral recognition through neural networks in MATLAB. It also explains the scope for further research in the area.