Related Experiment Video
Updated: May 25, 2025
08:05
Design and Analysis for Fall Detection System Simplification
Published on: April 6, 2020
10.6K
A Sliding Window-Based CNN-BiGRU Approach for Human Skeletal Pose Estimation Using mmWave Radar.
Yuquan Luo1, Yuqiang He1, Yaxin Li2
1School of Electronic Information Engineering, Beihang University, Beijing 100191, China.
Sensors (Basel, Switzerland)
|February 26, 2025
Summary
This study introduces a low-cost, low-power millimeter-wave radar system for accurate skeletal joint localization. The system enhances human pose estimation by fusing spatial and temporal data from radar point clouds.
Area of Science:
- * Computer Vision and Machine Learning
- * Radar Signal Processing
- * Human-Computer Interaction
Background:
- * Accurate human pose estimation is crucial for various applications, but existing methods face challenges with cost, power consumption, and data sparsity.
- * Millimeter-wave (mmWave) radar offers a promising alternative due to its ability to penetrate obscurants and provide dense point cloud data.
Purpose of the Study:
- * To develop a low-cost, low-power mmWave radar system for precise skeletal joint localization.
- * To enhance the accuracy and robustness of human pose estimation by effectively utilizing spatio-temporal information from radar data.
Main Methods:
- * Utilized a self-developed BHYY_MMW6044 59-64 GHz mmWave radar device for high-quality point cloud generation.
- * Implemented a sliding window mechanism to create multi-frame time-series data from single-frame point clouds.
- * Employed Convolutional Neural Networks (CNNs) for spatial feature extraction and Bidirectional Gated Recurrent Units (BiGRU) for temporal modeling, combined in a spatio-temporal fusion framework.
Main Results:
- * The proposed system accurately detects 25 skeletal joints, demonstrating significant improvements in positioning accuracy for fine joints like the wrist, thumb, and fingertip.
- * The spatio-temporal fusion framework effectively addressed the sparsity issue inherent in radar point clouds.
- * Achieved enhanced accuracy and robustness in pose estimation compared to existing methods.
Conclusions:
- * The developed mmWave radar system offers a viable low-cost, low-power solution for high-accuracy skeletal joint localization.
- * The spatio-temporal information fusion approach significantly improves pose estimation performance, particularly for subtle human movements.
- * The system shows strong potential for widespread adoption in human-computer interaction, intelligent monitoring, and motion analysis applications.
