TI AWR1843 Radar-Based 2D & 3D Human Activity Dataset for Six Action Classes

This dataset presents six-class human activity data captured using the Texas Instruments (TI) AWR1843 mmWave radar sensor. It includes synchronized 2D radar projection images and 3D point cloud representations for the following activities: standing, raising an arm, lifting a leg, bowing, squatting, and swinging arms. The data were collected under controlled indoor conditions with varying human subjects to ensure diversity in movement patterns and body dimensions. The radar captures reflected signals from human targets in real time, which are then processed using signal processing and point cloud reconstruction algorithms to generate spatial and temporal features. The dataset is designed to support research in radar-based human activity recognition, gesture analysis, and real-time monitoring applications. All samples are time-stamped, labeled by activity class, and stored in standard formats (.png for 2D projections and .pcd/.txt/.mat for 3D point clouds). This dataset offers a valuable benchmark for developing machine learning, deep learning, and signal processing algorithms for mmWave radar-based action recognition systems.本数据集包含由德州仪器（Texas Instruments，TI）AWR1843毫米波（mmWave）雷达传感器采集的六类人体活动数据。其涵盖针对以下六类活动的同步二维雷达投影图像与三维点云表征：站立、抬手、抬腿、鞠躬、下蹲及挥臂。数据采集于受控室内环境，招募了不同体型、运动模式各异的受试者，以确保数据的多样性。 雷达实时采集人体目标的反射信号，随后通过信号处理与点云重构算法进行处理，以生成时空特征。本数据集旨在为基于毫米波雷达的人体活动识别、手势分析及实时监测相关研究提供支撑。 所有样本均带有时间戳，按活动类别标注，并以标准格式存储：二维投影图像采用.png格式，三维点云则采用.pcd、.txt或.mat格式。本数据集可为基于毫米波雷达的动作识别系统的机器学习、深度学习及信号处理算法研发提供极具价值的基准测试数据集。

搜集汇总

背景与挑战

背景概述

该数据集基于TI AWR1843毫米波雷达传感器，捕获了六类人类活动（站立、举臂、抬腿、鞠躬、下蹲和摆臂）的同步2D雷达投影图像和3D点云数据，旨在为雷达基人类活动识别和实时监控研究提供基准。数据在受控室内环境下收集，涉及不同受试者以确保运动模式和身体尺寸的多样性，并经过信号处理和点云重建算法处理，以标准格式存储，支持机器学习和深度学习算法的开发。