以機器學習法分析線上簽名之動態特徵對於身分辨識影響之研究

• Main Author: 李宣儒
• Other Authors: 區國良
• Format: Others
• Language: zh-TW
• Published: 2013
• Online Access: http://ndltd.ncl.edu.tw/handle/77434012356001742557

碩士 === 國立新竹教育大學 === 數位學習科技研究所 === 102 === Identification plays an important role in the P2B environment. However, digital documents of online transactions can easily be forged by fake identities. Therefore, the commercial websites may have difficulty in guaranteeing the information security of all transactions. In contrast, banks have employed the off-line handwritten signatures many years when identifying the credit card owner. Therefore, developing a method for identifying the handwritten signatures on-line is the key to improve the information security of on-line commercial transactions. The on-line signature patterns and off-line signature patterns were extracted automatically when users were signing in the digital pen. At the same time, the machine learning methodology such as the Decision Tree, Bayesian Classifier and Support Vector Machine were employed for identification and detecting the forged signatures, which can improve the information security of on-line Chinese handwritten signatures. The goal of the study is to analysis the accuracy and security of identification using On-line Chinese handwriting characters. According to the experiment results, identifying the characteristics of the on-line Chinese handwriting characters recognition improved recognition accuracy by 6%, hence the use of on-line Chinese handwriting characters can fully identify forged signatures, which results in the improved safety of the user.