Font Size: a A A

Research And Implementation Of On-line Handwritten Signature Verification With Dyadic Wavelet Transform

Posted on:2015-02-25Degree:MasterType:Thesis
Country:ChinaCandidate:X Y LiuFull Text:PDF
GTID:2348330461480350Subject:Computer application technology
Abstract/Summary:PDF Full Text Request
With the rapid development of internet, it brings a huge change to the life and the work of people, in the meantime people adapt to it, they also use its advantage to acquire convenient for work and life. E-commerce ensues, but at the same time information security becomes the hot topic on which the people increasingly focus. Traditional identity verification is based on the items which are used for identity verification, but there maybe a stolen or a lost, its security is not perfect, users feel uncomfortable and inconvenient when they use it. Then identity verification based on biological characteristics comes into the sight of people.On-line handwritten verification is a common method based on biological characteristics. It has the characteristics of strong practicality and simple operation which is different from other verification with biological characteristics, this method plays an important role in the verification field for its uniqueness and stability. For today's rapid development of global e-commerce, on-line handwritten verification plays an important role in promoting it, and the research on it also has value and practical significance.This paper proposes on-line handwritten signature verification based on dyadic wavelet transform. Firstly, connect tablet with computer for acquiring the sample data of signatures, and the data is stored in the computer in the form of text documents. The collected data is preprocessed to remove interference and reduce workload, which are used for the noise reduction process mainly including pen up processing, smoothing, and normalization. They can reduce the workload for the following process of feature extraction. Secondly, extract the feature of the data which has been preprocessed through dyadic wavelet transform. Research and analyze the signal images from different scales and frequency through dyadic wavelet transform, determining a frequency and a scale which are helpful to identify true and false signatures, selecting the effective feature to reconstruct the new feature. Finally, acquire the dissimilarity between signatures which has been extracted the feature after wavelet transform by DTW algorithm. According to the mathematical statistics method, set the dynamic threshold to design the classifier in order to identify the signatures. Through a lot of experiments to prove, it can reach the FRR of 4%, and achieve a low error rate of 2%. The experimental results shows that the contents on which this paper researches can verify the signatures effectively.Then integrate the concept of cancellable signature into online handwritten signature verification. The main idea is to deform the signature images with some method, making users can only see the images after deformation, and the real signature images can not be contacted with the outside. If the database leaks to the external or someone attacks on the modules of the database maliciously, the cancel ability of this system will also to minimize the harm of users. It has played an important role to the users'personal information security, and can also protect the user's personal privacy. At the meantime, the experimental results of on-line handwritten signature verification based on cancelable signatures can reach the FRR of 8% and the error rate of 5%. Although its error rate compared with on-line handwritten signature a verification without the base of cancelable signature has a small range of ascension, it can not only identity signatures effectively, but also strengthen the security of information.
Keywords/Search Tags:Handwritten signature verification, Wavelet transform, Feature extraction, Threshold
PDF Full Text Request
Related items