基于Kinect的实时人体姿势识别

Kinect的实时骨骼跟踪技术获取身体关节点的三维位置信息,为进行人体姿势识别提供了丰富的信息,拟在三维关节点位置信息的基础上,建立一种实时的人体姿势识别系统。选择关节角度作为姿势特征,结合逻辑回归（logistic regression,LR）分类算法对54种姿势进行识别研究。实验结果表明,该姿势识别系统可以准确实时地识别人体姿势。

Real-time human posture recognition based on Kinect

Skeletal tracking provided by Kinect can acquire three-dimensional（3D） position of body joints,which provided a wealth of information for human posture recognition.A real-time human posture recognition system was built based on the 3D position of body joints.Joint-angles were selected as features and logistic regression（LR） algorithm was used to recognize 54 categories of postures.Experimental results showed that the human posture recognition system can recognize human postures in real time accurately.