Newspaper text recognition of Gurumukhiscript using and om forest classifier,

Abstract

Newspapers consist of very crucial information related to current as well memorable events. So, newspaper text needs to be preserved in a computer processable form for indexing of headline or making possible the search operations on newspaper text. For accurate results of recognition of text, appropriate classification of text based on extracted features is very important. Random Forest classifier is a widely used classifier in the field of pattern recognition and computer vision applications. In this paper, we have presented the recognition results using random forest classifier for newspaper text printed in Gurumukhi script. Different kinds of feature extraction techniques are used to extract the feature of characters that are fed to the random forest classifier. Standard k-fold cross validation and dataset partitioning strategy has been used for experimental work. Using the proposed method, maximum recognition accuracy of 96.9% and 96.4% has been achieved, using 5-fold cross validation and dataset partitioning strategy, respectively