FEATURE POINT BASED OFF-LINE CHINESE CHARACTER RECOGNITION SYSTEM

• Main Authors: Lin Tin-Hun, 林汀恆
• Other Authors: Wang Shuenn-Shyang
• Format: Others
• Language: zh-TW
• Published: 1998
• Online Access: http://ndltd.ncl.edu.tw/handle/01557447269295510999

碩士 === 大同工學院 === 電機工程研究所 === 86 === In the time of high technique, it is unavoidable to apply computers. Text input is the most important issue for the communication between human and machine, so the variant ways of text input are the topics of future research. The present developments of text input can be divided into four parts: (1) keyboard input (2) pen input (3) optical character recognition (4) voice input. Keyboard input is the most famous way to input text, but the problem is that we must be will trained or we can't get the faster input speed. Pen input is the same with our writing, easy to use, but the input speed is too slow. The OCR system is suit able for extracting available document, and it helps with the office automation. Voice input is not practical because of its low recognition rate. Our thesis is devoted to the research of the optical Chinese character recognition (OCCR). In this thesis, it is intended to investigate the recognition rate of thefeature point based Chinese character classifier. The feature points consist of the terminal points and fork junction points composed of strokes. We utilize the statistical feature of extracting the position and type of feature points to make reference patterns, and use the patterns to construct the reference database. After we compare with the input pattern and reference database, we output the character with the minimum error.