Human hand motion data collected by commercial motion capture system and custom-made data glove

This dataset presents diverse human hand motions in various forms including:Forward kinematic hand model implemented in 3d computer graphics software tool (/Joint angles/HandModel.blend);Bone lengths of 13 subjects (9 males, 4 females);Motion-captured rotations of the finger bones in Euler angles;Joint angles of the finger joints that were calculated by post-processing the motion-captured rotations.Also, with the aim of real-time tracking of these hand motions using the custom-made data glove, we provide:Sensor signals measured by the data glove during the hand motions;Regression model that relates the bone lengths and the sensor signals measured at the zero pose for instantaneous estimation of the bone lengths of the wearer;Filtering process for further refinement of the bone lengths estimation using the subsequent sensor signals;Filtering process for estimation of the joint angles using the subsequent sensor signals.ReadmeThe dataset is composed of three folders (/Initial bone lengths, /Refined bone lengths, and /Joint angles) and in each folder there exists Matlab or Python script that infers the collected motion-capture data and sensor signals in the /Data folder.How to use:Initial bone lengths: Run InitialBoneLengthsEstimation.m Refined bone lengths: Run RefinedBoneLengthsEstimation.m Joint angles 1. Run Calibration.py to output the linear model (already included in ./Data/230703_1440/model) 2. Open HandModel.blend, Open Text editor, and Press Run script (Alt+P) 3. Run JointAngleEstimation.py Detailed results, experimental procedures, post-processing methods of the data, and the filtering algorithms are described in a research article under review (5/2/2024).

本数据集涵盖多种形式的多样化人类手部运动数据，具体包含以下内容：
1. 基于三维计算机图形软件工具实现的正向运动学（Forward Kinematic）手部模型（文件路径：/Joint angles/HandModel.blend）；
2. 13名受试者（9名男性、4名女性）的手部骨骼长度数据；
3. 以欧拉角（Euler Angles）格式记录的指骨动作捕捉旋转数据；
4. 通过对动作捕捉旋转数据进行后处理计算得到的手指关节关节角数据。

为实现通过定制数据手套（data glove）对上述手部运动进行实时追踪，本数据集额外提供以下内容：
1. 手部运动过程中数据手套采集的传感器信号；
2. 关联骨骼长度与零位姿态下采集的传感器信号的回归模型，用于实时估算佩戴者的手部骨骼长度；
3. 利用后续采集的传感器信号进一步优化骨骼长度估算结果的滤波处理流程；
4. 利用后续采集的传感器信号估算手指关节角的滤波处理流程。

### 数据集说明
本数据集包含三个文件夹：/Initial bone lengths、/Refined bone lengths 以及 /Joint angles，每个文件夹中均配有Matlab或Python脚本，用于对/Data文件夹中采集的动作捕捉数据与传感器信号进行推断处理。

### 使用方法
1. 初始骨骼长度估算：运行脚本InitialBoneLengthsEstimation.m
2. 优化后骨骼长度估算：运行脚本RefinedBoneLengthsEstimation.m
3. 关节角估算：
(1) 运行Calibration.py以输出线性模型（该模型已预置在./Data/230703_1440/model路径下）；
(2) 打开HandModel.blend文件，进入文本编辑器，按下Alt+P快捷键运行脚本；
(3) 运行JointAngleEstimation.py。

本数据集的详细实验结果、实验流程、数据后处理方法以及滤波算法已收录于2024年5月2日提交的待审研究论文中。