基於Leap Motion之三維手寫中文簽名確認

Three- dimensional Chinese Signature Verification with using Leap Motion Controller

朱啓文(Zhu Qiwen)

指導教授 ： 范國清

國立中央大學/資訊電機學院/資訊工程學系/碩士(2015年)

摘要

簽名確認是一種以簽名作為生物特徵，用來確認簽名者身分的認證系統，已被廣泛用於許多場合。傳統的簽名確認系統因為數位化網路世代來臨，已不敷使用，新發展的簽名確認系統難以反應資訊科技的環境變遷，但仍多屬於接觸式輸入，必須借助物品或是碰觸感測器才能得到資訊，而造成許多不方便。本論文利用Leap Motion研究發展一個三維空中手寫簽名認證系統，在無需碰觸物品下，即可完成簽名確認，並同時具有安全性和方便性。

本論文利用Leap Motion擷得使用者簽名時的手部幾何及運動資訊，針對使用者指尖運動的簽名軌跡切割分段成實筆畫與虛筆畫，再設計抽取指尖動態資訊與手部幾何資訊中所隱含的三類特徵：行為速度、生物資訊和外型軌跡，作為簽名識別的依據。將真實簽名與仿造簽名樣本，藉由支持向量機器(Support Vector Machine, SVM)訓練出參考模型，用於判斷測試樣本的真偽。在實驗階段，本論文嘗試找出效能較佳的SVM分類器及核心函數，同時也試驗了模型訓練時，採用不同數量的仿造簽名負樣本，對分類器正確性的影響。論文中也實驗了不同特徵組合的區別能力。實驗結果顯示本論文所提出的特徵設計與確認方法，在簽名確認系統中具有良好的鑑別效果，提高仿簽偽冒者的被識破風險。

關鍵字

Leap Motion ； 中文簽名 ； 手寫簽名

並列摘要

Signature verification is kind of biometrics verification which can verify the identity of the signer using signature the feature as the biometrics feature. Currently, it has already been applied to many occasions. Because of security, the traditional signature verification system is not used. Besides, the current Signature verification almost is the contact type, it must be with the aid of the object or touch sensor to get information. It is not convenient. In the paper, we implement a 3D Signature verification system that has safety and convenience with Leap Motion, the signer can complete signature verification without touching any sensors or items.
In this paper, we get the hand information of signer’s signature by Leap Motion, such as fingers, and hands. The user’s signature tracks are divided into some strokes decided real strokes or virtual strokes. In the extraction phase, the features are extracted from the dynamic information, real strokes and virtual strokes, and the features are divided into three categories: speed, biology and appearance. By SVM, we train reference sample with true signature samples and counterfeit signature to improve the reliability of the system. Finally, the test samples and the reference samples are compared with the SVM. In the experiment, we choose better performance of SVM classifier and kernel function; decide to join how many fake signatures as negative samples to reference sample to improve system performance; and depend on the feature combination different judge which features have better ability to distinguish. Experimental results show that the various feature combinations toward genuine signatures and fake signature have different ability, and this system has good performance, also shows that the capacity to use Leap Motion as a 3D signature verification system sensor.

參考文獻

[1] Khan, M.A.U., Niazi, M.K.K., Khan, M.A.."Velocity-Image Model for Online Signature Verification." IEEE Transactions on Image Processing 15.11 (2006): 3540-49.

[2] Shivram, A., Ramaiah, C., Govindaraju, V.."A hierarchical Bayesian approach to online writer identification." IET Biometrics 2.4 (2014): 191-98.

[3] Xuhua, Y., Furuhashi, T., Obata, K., Uchikawa, Y.. "Constructing a high performance signature verification system using a GA method." Second New Zealand International Two-Stream Conference on Proceedings Artificial Neural Networks and Expert Systems (1995): 170-73.

[5] Lee, J., Yoon, J.-S., Soh, J.S., Chun, B.-T., Chung, Y.-K.." Using geometric extrema for segment-to-segment characteristics comparison in online signature verification." Pattern Recognition 37.1 (2004): 93-103.

[7] Liu, K., Huang, Y.S., Suen, C.Y.."Identification of fork points on the skeletons of handwritten Chinese characters." IEEE Transactions on Pattern Analysis and Machine Intelligence 21.10 (1999): 1095-100.