Related Experiment Video
Updated: Aug 12, 2026
14:27
Identification of Disease-related Spatial Covariance Patterns using Neuroimaging Data
Published on: June 26, 2013
A Spatial Millimetre-Wave Radar Point Cloud Dataset for Health Applications
Kailu Guo1, Elif Dogu2, Khalid Z Rajab3
1School of Electronic Engineering and Computer Science, Queen Mary University of London, London, E1 4NS, United Kingdom. kailu.guo@qmul.ac.uk.
Scientific Data
|August 10, 2026
Summary
This study introduces a new millimetre-wave radar dataset for analyzing human movements, crucial for developing advanced radar-based motion analysis systems. The data supports applications in daily activities and rehabilitation, enhancing human-computer interaction.
Area of Science:
- Computer Vision
- Robotics
- Biomedical Engineering
Background:
- Human motion analysis is critical for applications ranging from healthcare to human-computer interaction.
- Radar technology offers a privacy-preserving alternative to vision-based systems for motion capture.
- Existing datasets often lack the detailed annotations or controlled conditions necessary for robust radar-based analysis.
Purpose of the Study:
- To present a novel, controlled millimetre-wave radar point-cloud dataset.
- To facilitate research in radar-based human motion analysis for daily and rehabilitation activities.
- To provide synchronized radar data with 3D skeleton labels for supervised learning.
Main Methods:
- Collected millimetre-wave radar point clouds and Azure Kinect 3D skeleton data from 26 participants.
- Participants performed guided daily and rehabilitation movements under normal and low-light conditions.
- Processed and aligned radar frames with corresponding skeleton labels, annotating each frame with action and subject ID.
Main Results:
- The dataset contains approximately 3.47 minutes of quality-controlled movement data per participant per lighting condition.
- Includes 137,717 paired radar-skeleton frames with detailed annotations.
- Demonstrates the feasibility of using radar point clouds for action classification, identity recognition, and keypoint detection.
Conclusions:
- The presented dataset is a valuable resource for advancing radar-based human motion analysis.
- Enables the development and validation of algorithms for diverse applications, including rehabilitation and human-computer interaction.
- The controlled lighting conditions enhance the robustness and generalizability of learned models.
