基于运动学和动力学参数的步态识别研究

为了通过步态进行身份鉴别 ,采用模式识别的方法进行基于步态运动学和动力学参数的步态识别 ,提出了对步态运动学参数中关节角度、动力学参数中地面垂直支反力 (VGRF)进行特征提取的方法。对 14名健康男性受试者的步态进行了识别实验。结果显示 ,基于关节角度的识别准确率 (CCR)为 87.1 (k-近邻法中 k=1)和90 (k=3) ;基于 VGRF的 CCR为 85 .7 (k=1)和 80 (k=3)。另外 ,从多元统计学的角度对实验数据进行分析证明了通过步态进行身份鉴别的可行性。同时 ,主成分分析显示 ,在 VGRF曲线中 ,对步态识别起比较重要作用的是足部接触地面瞬间的部分。

A Study of Gait Recognition Based on Kinematics and Kinetics Parameters

In order to recognize people by their gait, we propose a pattern recognizing method based on kinematics and kinetics parameters. The feature extraction methods of joint angle and vertical ground reaction force (VGRF) were given. 14 healthy male subjects participated in this experiment. The experimental results showed that the correct classification rates (CCR) was 87.1% at k=1 and 90% at k=3 (k-nearest neighbor) based on joint angle recognition; the CCR was 85.7% at k=1 and 80% at k=3 based on VGRF recognition. The multivariate analysis of the experimental data proved the feasibility of gait recognition. The principal component analysis and curves of VGRF also showed that the instant of foot strike plays an important role in gait recognition.