mmWave-3DPCHM-1.0：A 3D Point Cloud Dataset for Human Actions Using Millimeter-Wave Radar

This dataset is a three-dimensional point cloud dataset of human motion using millimeter wave radar (referred to as mmWave-3DPCHM-1.0), which includes point cloud data collected and processed by two types of millimeter wave radar sensors. The two data acquisition devices are TI's IWR1443-ISK and Vayyar's vBlu RF imaging module. This dataset includes raw point cloud data from millimeter wave radar and preprocessed point cloud data, as well as point cloud processing codes, images and videos of various predetermined actions. This dataset is collected in a conference room. The data collection process involves the use of computers, brackets, and millimeter wave radars. The radar is installed on the wall at a height of 1.75 meters above the ground. Volunteers complete prescribed actions within a rectangular area, which is 1.2 meters horizontally from the radar and 2 meters in length and width. This dataset selected 7 volunteers and collected 12 types of human movements, including 3 static movements (falling, sitting, and standing) and 9 dynamic movements (punching, jumping, waving left, leaning forward left, opening arms, waving right, leaning forward right, squatting, and walking). Collect 3 sets of data for each action, with each set lasting 3 minutes. During the collection process, volunteers continuously complete the prescribed actions. Save the data as an Excel file. For the convenience of researchers, we preprocessed the data of TI millimeter wave radar by multi frame fusion clustering, etc. The processed code and results are all in the corresponding folder. Data paper and citation format: [1] Jin Biao, Sun Kangsheng, Wu Hao, etc A dataset and method for human motion recognition based on millimeter wave radar 3D point cloud [J]. Journal of Radar (in Chinese and English), 2025, 14(1): 73–89. doi: 10.12000/JR24195 JIN Biao, SUN Kangsheng, WU Hao, et al. 3D point cloud from millimeter-wave radar for human action recognition: dataset and method[J]. Journal of Radars, 2025, 14(1): 73–89. doi: 10.12000/JR24195

本数据集为毫米波雷达（millimeter wave radar）人体运动三维点云数据集（mmWave-3DPCHM-1.0），包含两类毫米波雷达传感器采集并经处理后的点云数据。两款数据采集设备分别为德州仪器（Texas Instruments, TI）的IWR1443-ISK模块与Vayyar公司的vBlu射频成像模块。
本数据集涵盖毫米波雷达原始点云数据与预处理点云数据，同时附带点云处理代码、各类预设动作的图像与视频素材。
本数据集采集于一间会议室中，数据采集过程用到计算机、支架与毫米波雷达设备。雷达安装于距地面高度1.75米的墙面，志愿者需在距雷达水平距离1.2米、长宽均为2米的矩形区域内完成指定动作。
本次数据集共招募7名志愿者，采集了12类人体动作数据，其中包含3类静态动作（跌倒、坐姿、站姿）与9类动态动作（出拳、跳跃、左挥手、左前倾、张开双臂、右挥手、右前倾、下蹲、行走）。每个动作采集3组数据，单组时长为3分钟，采集过程中志愿者持续完成预设动作。数据以Excel文件格式存储。
为方便研究人员使用，团队通过多帧融合聚类等方法对TI毫米波雷达的原始数据进行了预处理，处理代码与结果均存放于对应文件夹中。
数据集相关论文及引用格式如下：
[1] 金彪, 孙康胜, 吴昊, 等. 基于毫米波雷达三维点云的人体动作识别数据集与方法[J]. 雷达学报, 2025, 14(1): 73–89. doi: 10.12000/JR24195
JIN Biao, SUN Kangsheng, WU Hao, et al. 3D point cloud from millimeter-wave radar for human action recognition: dataset and method[J]. Journal of Radars, 2025, 14(1): 73–89. doi: 10.12000/JR24195

搜集汇总

背景与挑战

背景概述

该数据集是一个基于毫米波雷达的三维点云数据集，专门用于人类动作识别研究，包含两种雷达传感器采集的原始和预处理点云数据。数据集涵盖了12种人类动作（包括静态和动态），由7名志愿者在受控环境中完成，并提供了数据处理代码和相关文档，适用于动作识别算法的开发和验证。