基于MEMS传感器数据拓扑分析的用户识别

User Identification Based on MEMS Sensor Data Topology Analysis

随着科技的迅速发展,智能手机安全性的问题越来越受人们关注,用户识别认证在保证智能手机安全方面扮演着重要的角色。提出了一种基于微机电系统(Microelectro Mechanical Systems,MEMS)传感器步态信号持续同调的用户识别的新方法,MEMS传感器信号中的混沌不变量(伪周期性)可以由相空间中的拓扑特征来表示,通过持续同调提取相空间拓扑特征,建立不同用户的持续同调模型。利用持续同调特点,进一步简化持续同调为概率密度,并使用K-L散度度量不同用户模型间的差异,最终利用MEMS传感器信号实现无须用户主动参与的身份认证。为验证持续同调模型,采用十个不同用户,两种运动状态数据,通过仿真完成用户识别,结果表明识别的平均准确率达到95.3%。

With the sharp progression of technology, the security of smart phone draws more attention, in whichuser identification authentication plays a vital role. This study proposed a novel method of user identification based onpersistent homology of gait signals by Microelectro Mechanical Systems( MEMS). Homological characteristics ofphrase space representing chaotic invariant(pseudo periodic) in MEMS sensor signals, were extracted by persistenthomology to establish persistent homology models of users. Persistent homology was further simplified into probabilitydensity based on characteristics of persistent homology. Kullback-Leibler divergence was used to measure disparitiesin user models. MEMS sensor signals were used to realize the identity authentication without active participation of us-ers. To verify models of persistent homology, the data of ten different users in two kinds of states of motion were usedto complete user identifications, of which the average accuracy is 95. 3%.