基于相对运动特征的步态识别方法

Gait Recognition Based-on Relative Motion Features and Kinect

步态识别是一项很重要的生物识别技术。步态特征分为静态与运动两类。一些工作发现,尽管步态的本质特性在于运动,但运动特征的识别能力较为有限。究竟是运动本身的区分能力有限,还是现有特征未能充分捕获运动特性目前尚不清楚。基于Kinect骨骼模型,相对运动特征被提出,特定关节点之间的距离及其变化被用来刻画步态。相对运动特征的获取无需计算步态周期,鲁棒性好。实验结果显示,相对运动特征的识别精度达84%,与静态特征相当;与已有运动特征相比,精度提高10%以上。

Gait recognition Kinect is a very important biometric technology. Gait feature can be divided into two categories: static and dynamic. Some work found that although the essential characteristics of gait are motion, but the motion feature's recognition capability is limited. However, is the distinguishing ability of the motion itself limited, or existing features doesn't adequately capture the characteristics of motion? Relative distance-based motion features were proposed, and distances between the particular skeleton points and their changes were used to characterize the gait. The relative motion features were extracted without calculating the gait cycle, and with robustness. Experimental results show that the relative motion features' recognition accuracy is up to 84%, which is equivalent to static characteristic.