Kinect v2 skeleton joints point and 6 angles dataset

The collected dataset for eight fixed positions came from 127 volunteers (men and women), whose ages ranged from 20 to 41, with different heights (1.45–1.91 m) and different body sizes.
Each person among the volunteers imitates or performs the eight positions or poses: “hands up”, “right hand up”, “left hand up”, “hands-on head”, “arms open”, “stand up straight”, “hands-on waist" and “hands forward” as shown in picture (1), interspersed with a simple movement that falls under the same position.
For each person, record 240 frames (each frame contains 15 joint camera coordinates in X, Y, and Z, and 6 angles as show in picture (3)). The record total frames are 30480 frames. 72% are used as training data and 28% are used for testing

本次采集的8种固定姿态数据集，来自127名男女志愿者，其年龄跨度为20至41岁，身高范围1.45–1.91米，且体型各异。每名志愿者需模仿或完成以下8种姿态：“双手上举”“右手上举”“左手上举”“双手抱头”“双臂张开”“挺直站立”“双手叉腰”及“双手前伸”，动作展示如图1所示，期间穿插属于同一姿态类别的简易过渡动作。针对每名志愿者，采集240帧数据，每帧包含15个关节点的X、Y、Z三维相机坐标，以及6个角度参数，详情如图3所示。本次采集总帧数为30480帧，其中72%用作训练集，剩余28%用作测试集。