Ancient Tamil Character Recognition using Ensemble Classifier

Ancient Tamil character recognition of the stone manuscripts are rich in the source of art, music, literature, management and administration. It therefore becomes critical to access those stone manuscripts belong to 12th century period by the epigraphers to share the contents present in the manuscripts with the people. In this paper the Ensemble based classification of OCR (Optical Character Recognition) is proposed to give more accuracy of recognizing those characters inscribed on the stone manuscript. The testing sample is preprocessed to remove noises present in the inscriptions using Median filtering. The noise free images are segmented using Bounding Box technology to segment each character. The global and local features are extracted from each character i.e 79 features for each character in the dataset. The one third of the extracted features are applied to the Bagging based KNN and Discriminant and also Boosting technique. The Bagging based technique correctly classifies the characters among 66 classes and the correctly classified character is mapped to Unicode of the Modern Tamil character. The performance measure of Segmentation rate and Recognition rate is calculated.