Biometric Gait User-identification Base On Accelerometer

With the development of society,information is getting closer to personal life.Today's society is facing the security challenges like information's leakage and steal,high-tech crime,terrorist's attack,and so on.Society needs a new authentication technology to solve these problems.Because of the advantages of universality and uniqueness of Biometric identification technology,it has received more and more attention.Meanwhile gait is human behavior characteristics and attracts many scholars study.User-identification using gait is a new direction in the study and has great research value.This paper studied biometric gait user-identification based on accelerometer.The characteristics of acceleration sensor's high penetration and the signal's easy acquisition bring great convenience to gait identification research.This article uses the gait acceleration signal,by identifying the gait acceleration signal to achieve the goal of recognition gait gatherers individuals.In this paper,the main work is as follows:This paper have done detailed research on the gait of the acceleration signal data processing and designed the area for maximum cycle split points.This paper explores the gait identification effect using DTW based on typical cycle.Through the experiment,the x axis acceleration signal's best experimental results: FRR=2.56%,FAR=3.21%;the y axis acceleration signal's best experimental results: FRR=5.98%,FAR=5.63%;the z axis acceleration signal's best experimental results;FRR=6.84%,FAR=7.69%.By comparison with experimental observation found that the x axis acceleration signal recognition effect is best,at the same time the highest stability.The integration of the x axis and y axis DTW distance for the further experiment,get the best results: FRR=0.85%,FAR=1.78%,the results are further improved.This paper innovation points are as follows:(1)Gait acceleration feature extraction use the method based on cycle extraction.In order to solve the problem of periodic error-prone,this paper puts forward the method looking for cycle dividing point through the maximum regional value.The method uses double limit of peak's height and width,then connecting with the period length to judge cycle dividing point.This method effectively solves the problem of the peak appearing double extreme value points and greatly reduces the cycle extraction error rate.In order to get rid of the bad period and data redundancy and reduce the time consumption,this paper design the methods of extracting typical period by the average cycle,greatly reduced the time complexity.(2)In order to solve the problem of acceleration sensor's different installation angle,this paper has designed a rotation DTW matching method.The method uses two typical gait cycle's rotation to find the angle between the two samples.In order to quickly find the angle difference between samples,this paper designs the dichotomy methods for greatly improving the efficiency.Using rotation DTW method based on typical gait cycle,the best results is FRR = 0.92% and FAR = 0.59%,the FAR has a bigger drop.