结合DTW与改进STLCS的动态手势认证研究

Research of dynamic gesture authentication based on DTW and improved STLCS

智能手机用户身份认证是保证手机信息安全的重要方法。为了有效利用人体动态手势局部变化特征并提高认证精确度,提出一种结合动态时间规整与改进短时最长公共子序列算法的动态手势认证方法。在第一级认证阶段利用基于欧式距离的动态时间规整算法进行手势加速度信号全局相似度度量,判断相似度分值是否处于容错区间内,引入短时滑动窗口用于获取手势局部变化特征并采用最长公共子序列算法进行第二级认证。实验结果表明,针对模仿动作攻击,等错率达到了1.11%,比已有的LCS模型降低了0.97%,比已有的DTW模型降低了1.91%,有效提高了认证精度,加强了认证系统抵御模仿动作攻击的能力。

Intelligent mobile phone user authentication is an important method to ensure the information security of mobile phone. In order to use the local characteristic of dynamic gesture and improve authentication accuracy,a dynamic gesture authentication method based on Dynamic Time Warping( DTW) and improved Short Time Longest Common Subsequence( STLCS) is proposed in the paper. In the first level authentication stage an algorithm called Euclidean Dynamic Time Warping is used to measure the similarity of global gesture acceleration signals,then the judgment about whether the similarity score is in the fault-tolerant interval is done. The short time sliding window is introduced to get the local characteristics of gestures and the second level authentication is carried out by longest common subsequence algorithm. The experimental results show that,in the case of imitation attack,Equal Error Rate( ERR) reaches 1. 11%. Compared with the existing LCS model,ERR has a decrease of 0. 97%. Compared with the existing DTW model,ERR has a decrease of 1. 91%. So this method improves the precision of authentication and strengthens the authentication system's ability to resist imitation attack effectively.