协同交互情境下双视图耦合的多用户身份识别

针对目前协同交互情境下,因遮挡和接触等因素引发的多人协作身份识别错误问题,提出一种双视图耦合的多用户身份识别方法.借助深度相机,分别通过骨骼运动跟踪和卡尔曼滤波建立双视图运动跟踪.利用双视图的用户运动跟踪数据构建相互关联的耦合有限状态机,对耦合关系中的具体运动状态进行分析,建立规则算法;引入实时正误标记值进行多用户身份识别的实时监控与耦合矫正.所提方法与基于单视图的多用户身份识别方法进行对比实验,结果表明,在协同交互情境下所提方法对多用户身份识别更具有鲁棒性.

Multi-user identification based on double views coupling in cooperative interaction scenarios

An approach of multi-user identification based on double views coupling was proposed for the problem of identification error in cooperative multiplayer caused by occlusion and contact, in cooperative interaction scenarios. Double views motion tracking was established by the method of skeletons motion tracking and Kalman filter, with the depth cameras. Correlative finite-state machine was build using the motion tracking data of double views, to analyze the specific user motion status of coupling relationship and set up algorithms, the value of true or false tag was introduced to monitor and coupling correct multi-user identity in real-time. The new approach was compared with that of multi-user identity based on single view by experimentation, it is indicated that, the approach of multi-user identification based on double views coupling in cooperative interaction scenarios is more robust.