Research On In-air Chinese Signature Authentication Based On Accelerometer

While people are more willing to shop and make electronic payment through mobile Internet,they have more pressing need for the information security of mobile Internet.Biometric authentication methods develop rapidly in recent years and they have been widely used in many fields.The in-air Chinese signature authentication method which makes use of both physiological and behavioral biological characteristics is considered to be safe and reliable.This paper proposes a new in-air Chinese signature authentication method.Signature sequences are segmented based on the extreme point matching results.Then a decision-level fusion method based on DTW,Hausdorff distance and discrete Fréchet distance which are used to measure similarity between different segments is applied.Main work of this paper is as follows:(1)This paper completes the data acquisition.Users are required to perform their in-air Chinese signatures while holding a circuit board with a three-axis accelerometer embedded.Original acceleration data of in-air Chinese signatures are processed by the differential threshold method to remove invalid signature data.Then in-air Chinese signatures should be preprocessed,involving data smoothing and length normalization.(2)Acceleration data of signatures are divided into several segments on the bases of the Chinese character structure.Method of segmentation in this paper is based on the matching result of extreme points in acceleration data curves.While applying the classical DTW algorithm to match extreme point sequences,a lot of extreme points are mismatched which means different kinds of extreme points are aligned.To obtain the correct matching result,this paper changes the limiting condition of classical DTW.Aiming to eliminate unintuitive alignments where a single extreme point on one sequence corresponding to a series of extreme points of another sequence,this paper merges some extreme points.As a result,extreme points of signatures are one-to-one corresponding and these signatures are segmented.(3)Similarities between different segments are measured.After applying DTW,Hausdorff distance and Fréchet distance to measure similarities between acceleration data segments,experimental results show that all of the three methods could improve the authentication performance to some extent.But each of the methods has advantages and disadvantages and represents part of the information of the user.To make full use of all of the information reflected by these three methods,this paper applies the information fusion method.Distance results obtained by the three methods are transferred to confidence.Then the weighted average method is applied to complete the information fusion.For the purpose of analyzing the performance of the method proposed in this paper,a database with 1360 in-air Chinese signatures collected from 34 users is built.These signatures involve 102 template signatures,578 genuine testing samples and 680 forged samples.Finally,after testing all of the in-air Chinese signatures in the database,a better authentication result of FRR: 4.15% and FAR: 3.38% is obtained.