線上中文簽名驗證系統

On-line Chinese Signature Verification System

黃俊欽(Chun-chin HUANG)

指導教授 ： 林啟芳

元智大學/工程學院/實業計劃研究所工學組/碩士(1992年)

摘要

辦公室自動化的發展使得以電子資訊如電子郵件(E-mail)，成為未來資訊
交換的重要媒介，而在傳遞電子郵件(E-mail)如何利用電腦來證明文件出
處的身份驗證問題就成為重要的工作。簽名驗證即用比對特定簽名筆跡的
方式來做身份驗證的工作。本論文即研究線上中文簽名驗證的方法。我們
利用數位板(tablet)來輸入簽名資料，由數位板可獲得簽名者簽 名的靜
態樣式(static patterns)（如影像）及動態樣式(dynamic patterns)（
如書寫速率）資料。在學習簽名中階段我們將同一簽名 者的三次簽名樣
式作比對處理，然後用其中較具代表性的簽名樣式來 建立參考簽名檔。
由於從數位板上獲得的簽名樣式，會因為機器或人為的因素而呈現不 是
很穩定的現象，我們提出手寫筆跡相似度(handwriting curve
similarity measurement)演算法來計算參考簽名及待測簽名的靜態 相似
度，然後削去靜態相似度較差的筆劃。 另外我們發展速率曲線 相似度
velocity curve similarity measurement)演算法來計算剩下筆劃的動態
相似度然後以這兩個相似度的值來判斷是否為正確的簽名。 最後我們再
以計算類Ⅰ錯誤 (type-Ⅰ error)及類Ⅱ錯誤(type-Ⅱ error)的方式來
做系統執行效率的評估，以說明我們所提方法的優越性。

關鍵字

簽名驗證 ； 樣式比對 ； 相似度

並列摘要

The personal identification via computers is more and more
important in the recent years. The reasons are due to the fast
development of office automation, and the need for information
interchange by computers. One of the most commonly used method
for identifying person is by verifying their signatures. The
purpose of this paper is to develop methods to solve the
problems of on-line chinese signature verification. The system
we proposed consists of the following two stages: the learning
stage and the test stage. In the leaning stage, the signer was
instructed to sign his name on a tablet three times. The
static signature patterns (e.g., the image data) and the
dynamic signature patterns (e.g., the velocity of hand
movements) can be obtained by sampling the signatures. We can
use these patterns to create a reference signature file by
comparing the three signatures and choosing the 'best' one. The
reference signature file will be used in the test stage.
Because signature patterns obtained by human and machine are
usually unstable. We present a handwriting curve similarity
measurement method to calculate the similarity between the
static patterns of the reference signature and that of the test
signature. Some of the signature strokes which are not similar
enough are discarded, and the dynamic velocity curve is created
for each of the remaining strokes. We also developed a velocity
curve similarity measurement method to calculate the similarity
for these strokes. The two similarity measurement values are
used to verify the correctness of the signature. Finally, the
performance of the proposed approach is evaluated by computing
the type-Ⅰand the type-Ⅱ error rates.

並列關鍵字

Signature verification ； Pattern matching ； Similarity