Dynamic Thresholding based Ancient Character Recognition using Enhanced Convolution Neural Network

  • P. Balasubramanie, E. K. Vellingiriraj, R. Logesh Babu, M. Dinakaran

Abstract

Olden character recognition has been the majorcomplicatedjob which will have diverse shapes as well asstructural formats. There isdifferentstudytechniques areestablishedpreviouslytowards accurate prediction of ancient characters. In our previous work, this is attained by introducing the method namely adaptive character recognition Method (ACRM). However, in this research work character segmentation is not done clearly which would reduce the recognition accuracy. It has been resolved within the projectedstudy work throughestablishing the techniqueknown asDynamic Thresholding based Ancient Character Recognition Method (DT-ACRM). In this work, initially gray scale conversion based prior processing has been done for removing the noise existsinto the images. After that accurate character segmentation is done by using dynamic thresholding method which would lead to accurate character recognition. Finally character recognition is performed by using enhanced convolution neural network. The performance evaluation of the studytask has beenmadeinto the matlab replicationsurroundingsalsothis hasconfirmed that the projected work provides enhanced solution than previous works.

Published
2020-03-19
How to Cite
R. Logesh Babu, M. Dinakaran, P. B. E. K. V. (2020). Dynamic Thresholding based Ancient Character Recognition using Enhanced Convolution Neural Network. International Journal of Advanced Science and Technology, 29(3), 5679 - 5689. Retrieved from http://sersc.org/journals/index.php/IJAST/article/view/6193
Section
Articles