Signature Verification Based On Video

Signature verification is an age-old technology in the field of personal authentication,but it coruscated new vitality under the conditions of modern computer technology,and it was become research focus to analyse signatures using the theory of pattern recognition.Generally,automatic signature verification can be divided into two areas depending on the data acquisition method:on-line and off-line signature verification.The traditional on-line and off-line methods have their advantages and disadvantages.In this article, we mainly do some research on video-based signature verification.Firstly,this article introduces the concepts of video-based signature verification,the mostly difference between this method and traditional on-line and off-line methods is that a camera is set to capture the whole process of signing.This method can be considered as a special on-line method,it can not only get the dynamic information,but also static information and some other biometrics information,such as:gesture,fingernail and etc.Due to the using camera,locating and tracking the pen tip in video stream is a key step of video-based signature verification.This article presents a contour-based method to locate most kinds of pens tips based on observing the manners of most people handling the pen with their hand.And then a tracking method based on particle filter and template matching is proposed by summarizing the methods of tracking and considering the nonlinear movement of pen tip during signing and the need of locating the pen tip exactly on pixels level.Kinematic theory of rapid human movement is a kind of theory which describes basic movement of human body,and it is widely used in on-line signature verification.This article analyses the data in video-based signature verification,as a special on-line method,video-based signature still adapts for kinematic theory,but it has its own characteristics,such as its information is not rich and interval is so disperse.Based on these analysis,this article proposes a segmentation method of video-based signature based on kinematic theory of rapid human movement,and uses it in a classify method based on dynamic time warping of strokes.Finally,the experiments show that the method of locating pen tip proposed in this article is robust for most kinds of pens,the method of tracking pen tip proposed in this article is both efficient and accurate,the method of video-based signature segmentation is directive significance,but need to perfect in future.The recognition rate on self-built video-based signature database is good.