Suvdaa Batsuuri

Afiliation: Associate Professor at School of Engineering and Applied Sciences, National University of Mongolia

 

Abstract

Mongolian many historical and cultural documents are stored as books, whch are written in traditional mongolian script. Therefore, traditional script recognition is one of the important topics in Mongolia. In this paper, we have summarize results of segmentation of the traditional mongolian script. There are 3 methods are discussed; (1) handwritten or italic script segmentation by computing the slop, (2) segmentation by backbone width, and (3) segmentation by removing backbone. Specially, in the ‘modon bar’ format which is printing style without any separated parts called “dusal”. In experiments, we use “Ganjuur Danjuur sudar” cultural valuable book with 102 pages. In this time, we test 3 pages, 84 columns, 703 words, 2500 scripts. As a results, we achieved the column segmentation rate 100%, the word segmentation rate 98% and script segmentation rate 85%.

Short Bio

Suvdaa Batsuuri received bachelor and master’s degree from National University of Mongolia in 2002 and 2004, respectively. She got Ph.D in Computer Science at department of Computer and Software Engineering, Kumoh National Institute of Technology, South Korea, in 2011. From 2002 to 2007, she was working at National University of Mongolia. In 2010, she was a part-time lecturer at Kumoh National Institute of Technology. Since 2011, she is associate professor at department of Information and Computer Science, school of Engineering and Applied Sciences of NUM. Her research interests including image processing, computer vision, pattern recognition, distance metric learning and artificial Intelligence.