手势识别身份认证的连续隐马尔可夫模型

Continuous Hidden Markov Model for Hand Gesture Recognition

针对智能手机存储隐私信息所面临的安全问题,采用一种基于连续隐马尔科夫模型的手势识别身份认证方法.首先由手机触摸屏传感器采集手指滑动的原始手势特征序列,并通过大小归一化及平滑去噪预处理;接着提取手势运动轨迹的三个基本特征序列与三个隐含的特征序列;最后采用概率统计的方法,使用连续隐马尔科夫模型建立用户手势模型,用于测试比较特征序列以判断用户身份的合法性.仿真实验结果表明,与动态时间规整算法和支持向量机算法相比,方法具有较低的错误拒绝率和错误接受率,能明显提高身份认证的准确性.

Aiming at the security problem of storing privacy information in smart phone, a method of hand gesture recognition based on continuous Hidden Markov Model is proposed. Firstly, the original gesture sequence of the finger sliding is collected by the touch- screen sensor of the mobile phone,and the gesture sequence will be preprocessed by using normalization and smoothing. Then the fea- tures, which are three basic feature sequences of the hand movement trajectory and three hidden feature sequences, will be extracted from the preprocessed sequence. Finally, a statistical method that is continuous Hidden Markov Model is adopted to establish the user gesture model, which will be extracted to test the feature sequence to determine the legitimacy of the user＇s identity. The simulation re- sults show that the proposed method has lower error rejection rate and false acceptance rate than that of the Dynamic Time Warping al- gorithm and Support Vector Machine algorithm, and can significantly improve the accuracy of identity authentication.