การรู้จำตัวอักษรเขียนภาษาไทยที่เป็นคำแบบออฟไลน์โอยใช้หลักเกณฑ์ทางฟัซซีร่วมกับคุณลักษณะบ่งความต่าง / อิทธิพันธ์ เมธเศรษฐ = Off-line Thai word based handwritten character recognition using fuzzy rules with distinctive features / Ithipan Methasate
The objective of this thesis is to propose a method of Thai word based handwritten character recognition by applying fuzzy logic with distinctive features. The system considers the similarity of the handwritten character and character patterns of the 76 province names. The system chooses 5 candidates in each character to check with the lexicon in postprocessing step. In the lexicon checking step, the length grouping is used to reduce the processing time. The experiment was wexecuted on a microcomputer with PentiumIII 600 MHz and 128 Mbytes RAM. The test includes 100 patterns of 44 Thai consonant characters totalling 4400 characters; 20 patterns of vowel and tonal marks totalling 760 characters, and 40 patterbs of the 76 words totalling 23,720 characters. The error rate in character recognition is 3.62% and in word recognition the error rate is 1.71%.