基于无标记点运动跟踪的步态参数提取

提出一种无标记点的步态参数提取方法,采用基于模型的人体运动跟踪方法,无须标记点即可从视频图像序列中获取运动步态参数。利用粒子滤波算法跟踪人体目标运动过程,通过基于外观模型的相似度计算及边界匹配误差判断定位腿部关节点。依据曲线拟合方法在跟踪与定位结果上提取运动步态参数。实验结果表明,该方法能精确定位腿部关节点且有效地提取步态参数。

Extraction of gait parameters based on marker-less motion tracking

This paper proposed a novel approach for extraction of gait parameters from video image sequences without marker by human motion tracking based on human model.The approach tracked the process of human motion with particle filters,and located the joints of legs through likelihood computing with the appearance models and boundary matching error.Then extracted kinematical gait parameters from the result of tracking and locating by curve fitting.Experimental results show that the approach can locate the joints of legs accurately and capture the gait parameters effectively.