Line-Level Script Identification for Six Handwritten Scripts Using Texture Based Features

Abstract

Script identification from a given document image has some important applicability in many computer applications such as automatic archiving of multilingual documents, searching online archives of document images and for the selection of script specific Optical Character Recognition (OCR) engine in any multilingual environment. In this paper, we propose a texture based approach for text line-level script identification of six handwritten scripts namely, Bangla, Devnagari, Malayalam, Tamil, Telugu and Roman. A set of 80 features based on Gray Level Co-occurrence Matrix (GLCM) has been designed for the present work. Multi Layer Perceptron (MLP) is found to be the best classifier among a set of popular multiple classifiers which is then extensively tested by tuning different parameters. Finally, an accuracy of 95.67 % has been achieved on a dataset of 600 text lines using 3-fold cross validation with epoch size 1,500 of MLP classifier.