|Author:||Tong, Wing-him Eric|
|Title:||On-line signature verification|
|Subject:||Signatures (Writing) -- Data processing|
Optical character recognition devices
Optical pattern recognition
Hong Kong Polytechnic University -- Dissertations
Department of Electronic Engineering
|Pages:||vii, 70 leaves : ill. ; 30 cm|
|Abstract:||In this project, an on-line signature verification system is developed as a potential candidate of an automatic identification system. The potential application of the signature verification system is viewed as a security and identification measure in the area of the fast growing electronic or cyber communication world. The system consists of several components: data acquisition, pre-processing, segmentation, feature extraction, neural network training and verification. The verification system process starts from the data acquisition by a pen-based writing tablet input device together with an event driven program to trigger the signature capture. Raw data is pre-processed by a cleanup and normalisation procedure before segmentation. The signature trace is then segmented according to the natural signature speed, which is the unique characteristics of the signer. Timely and dynamic information is finally extracted and formatted prior to presentation for neural network training and verification. A back-error propagation three-layer MLP neural network with supervised training is selected for the purpose of verification and classification. Neural networks for both verification and classification are constructed during the training phase. Performance is evaluated after training and testing with different training sets. Some freehand forgeries are also used for evaluation. Experiment results show a high degree of overall accuracy and false acceptance rate for genuine signatures. However, the false reject rate is not satisfactory. The results for both verification and classification are similar without too much discrepancy. For future improvement, an additional component for segment alignment prior to training is suggested to align corresponding segments of the same signature. Also, a multistage neural network and/or combination classifier may be able to achieve better performance.|
|Rights:||All rights reserved|
Files in This Item:
|b14418897.pdf||For All Users (off-campus access for PolyU Staff & Students only)||2.39 MB||Adobe PDF||View/Open|
As a bona fide Library user, I declare that:
- I will abide by the rules and legal ordinances governing copyright regarding the use of the Database.
- I will use the Database for the purpose of my research or private study only and not for circulation or further reproduction or any other purpose.
- I agree to indemnify and hold the University harmless from and against any loss, damage, cost, liability or expenses arising from copyright infringement or unauthorized usage.
By downloading any item(s) listed above, you acknowledge that you have read and understood the copyright undertaking as stated above, and agree to be bound by all of its terms.
Please use this identifier to cite or link to this item: