On-line Handwritten Signature Verification Based On Speed And Pressure Partition

With the rapid development of network information, the network information security problems become more and more critical, so that requirement for security of network information is getting increasingly high. Biometric technology based on human biometric characteristics such as fingerprint, iris, signature and so on is widely used because of the advantages of its uniqueness, easy to carry, hard to be lost or forgotten and high stability. Signature is a kind of characteristic of act. On behalf of the signer’s unique information, it is one of the most widely-used biometric authentications. Compared to other biometric characteristics, it has the advantages of low-cost, difficult forgotten, easy to obtain, anti-counterfeiting, high acceptability and so on. On-line handwritten signature verification is a method that we identify the authenticity of the identity of the signer with a known person of test signatures specimen, to distinguish that this sample through a variety of technical and theoretical algorithm is real or fake.The feature selection and correct rate of online handwritten signature verification has been the focus of the research for online handwritten signature verification technology. Characteristic parameter method and characteristic function method need calculation of matching on large number of features, the training process contains complex computation and large storage. According to the experimental experience, speed and pressure is two features that can be directly collected and easy distinguished, because the forger can hardly imitate glyphs and the long-term habitual speed and pressure of the signer. According to this feature, this paper promotes a kind of online handwritten signature verification method based on the speed and pressure partition.Determining a base signature in registration signatures of each signer, utilizing the characteristics of high distinction of speed and pressure, and partition the base signatures. Builds templates for each partition, identify the most stable partition; find out the parameters of the partition to establish decision boundary. This algorithm reduces the calculated amount and storage capacity needed by the signature training, and improves the recognition efficiency of a signature.This paper firstly introduces the signature data acquisition process and pretreatment methods, which focuses on principal component analysis method to rotate the signature. Then it discusses in detail signature verification method based on the speed and pressure partition, and discuss the distribution of constant c in decision boundary and its impact of value on the verification results, through discussion to determine the optimal value of the constant c and calculate the corresponding FAR and FRR. The paper discusses the influence of partitioning with the x-direction speed instead of the speed of signature track to verify, and establishing double decision boundary to verify to the verification results, and points out that the proposed algorithm is optimal. Also introduces the dynamic time warping algorithm and BP neural network algorithm, the verification results of them and proposed algorithm, and introduces the comparative analysis of them.The experimental results show that determining different constant c value for different signatures, thereby balancing FAR and FRR, to reduce the average error rate, and to achieve high authentication rate above95%.