Related Experiment Video
Updated: Jul 9, 2025
06:25
Time Multiplexing Super Resolving Technique for Imaging from a Moving Platform
Published on: February 12, 2014
8.5K
Human Movement Recognition Based on 3D Point Cloud Spatiotemporal Information from Millimeter-Wave Radar
Xiaochao Dang1, Peng Jin1, Zhanjun Hao1
1College of Computer Science & Engineering, Northwest Normal University, Lanzhou 730070, China.
Sensors (Basel, Switzerland)
|December 9, 2023
Summary
This study introduces a new millimeter-wave radar system for accurate human movement recognition, achieving up to 94% accuracy for walking and 93% for falls. The system enhances privacy and works in challenging environments.
Area of Science:
- Computer Science
- Engineering
- Signal Processing
Background:
- Traditional human movement recognition methods like video and Wi-Fi struggle with performance in adverse conditions and complex environments.
- LiDAR-based methods are limited as they primarily capture static object characteristics, not dynamic human movements.
- Millimeter-wave radar offers advantages in privacy, security, and non-line-of-sight recognition for human motion analysis.
Purpose of the Study:
- To develop a novel human motion feature recognition system (PNHM) utilizing millimeter-wave radar 3D point cloud spatiotemporal information.
- To design and implement a neural network, based on PointNet++, for effective recognition of human motion features.
- To evaluate the system's performance in recognizing four distinct human motions across varied environments and angles.
Main Methods:
- A human motion feature recognition system (PNHM) was developed using millimeter-wave radar spatiotemporal 3D point cloud data.
- A neural network architecture, adapted from PointNet++, was designed for processing and classifying motion features.
- A dataset comprising four human movements (walking, squat-to-stand, stand-to-sit, falling) was created under controlled experimental conditions.
Main Results:
- The PNHM system achieved high recognition accuracies: 94% for walking upright, 84% for squatting to standing, 87% for standing to sitting, and 93% for falling.
- The system demonstrated robust performance across different angles and experimental environments.
- Comparison with four mainstream 3D point cloud action recognition models validated the proposed system's effectiveness.
Conclusions:
- Millimeter-wave radar 3D point cloud spatiotemporal information is effective for human movement recognition.
- The proposed PNHM system, leveraging PointNet++, provides a privacy-preserving and accurate solution for human motion analysis.
- This technology holds significant potential for applications in intelligent pensions, remote health monitoring, and child supervision.
