Please use this identifier to cite or link to this item: http://cmuir.cmu.ac.th/jspui/handle/6653943832/55531
Title: Lanna Dharma handwritten character recognition on palm leaves manuscript based on Wavelet transform
Authors: Papangkorn Inkeaw
Chutima Chueaphun
Jeerayut Chaijaruwanich
Atcharin Klomsae
Sanparith Marukatat
Keywords: Computer Science
Issue Date: 17-Feb-2016
Abstract: © 2015 IEEE. Lanna Dharma alphabet is used in the past in the North of Thailand, mainly for religious communication. Most of handwritten Lanna Dharma is found in form of old palm leaves manuscripts. These documents have not been properly preserved, still unprotected and damaged by the time. To preserve these valuable documents, handwritten optical character recognition is one of the first choices. This paper proposes an efficient method for Lanna Dharma handwritten character recognition from palm leaves manuscript image. In recent years, research towards Dharma Lanna character recognition from printed document is proposed. However, the proposed method cannot be applied to handwritten documents. This research aims to compare the different feature extraction methods for Lanna Dharma handwritten recognition. The first step in the proposed method is image preprocessing that binarized, enhanced, line segmented, level segmented and character segmented. The next step, each character image was extracted as feature vector using various feature extraction method based on Wavelet transform. Then several alternative feature extraction methods were compared by evaluating their effect on character recognition performance using K-Nearest Neighbor algorithm. The experimental results show that the best feature extraction is 2D, 1D wavelet transform and region properties feature extraction. The recognition rates of 10-fold crosses validation are 93.22 % for upper level, 95.48% for middle level, and 97.77% for lower level.
URI: https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=84971657237&origin=inward
http://cmuir.cmu.ac.th/jspui/handle/6653943832/55531
Appears in Collections:CMUL: Journal Articles

Files in This Item:
There are no files associated with this item.


Items in CMUIR are protected by copyright, with all rights reserved, unless otherwise indicated.