Related Experiment Video
Updated: Dec 20, 2025
06:37
Author Spotlight: Addressing Technical and Subjective Challenges in Measuring Classroom Attention
Published on: December 15, 2023
5.1K
Real-Time Human Action Recognition with a Low-Cost RGB Camera and Mobile Robot Platform
Junwoo Lee1, Bummo Ahn1,2
1Robotics Group, Korea Institute of Industrial Technology, Ansan 15588, Korea.
Sensors (Basel, Switzerland)
|May 23, 2020
Summary
This study presents a real-time human action recognition method using a single RGB camera, suitable for mobile robots. The approach achieves 70% accuracy on a benchmark dataset, running at 15 FPS on embedded systems.
Area of Science:
- Computer Vision
- Robotics
- Artificial Intelligence
Background:
- Human action recognition is crucial for applications like surveillance and robotics.
- Current deep learning methods using Kinect cameras are often too complex for mobile robot platforms.
- Existing models struggle with real-time performance on resource-constrained systems.
Purpose of the Study:
- To develop a real-time human action recognition system for mobile robots using only RGB cameras.
- To overcome the limitations of complex, depth-sensor-dependent models.
- To enable continuous human monitoring by mobile robotic platforms.
Main Methods:
- Integrated OpenPose and 3D-baseline libraries for skeleton joint extraction from RGB images.
- Employed convolutional neural networks for action classification.
- Utilized an NVIDIA JETSON XAVIER embedded board for real-time processing on a mobile robot platform with a tracking algorithm.
Main Results:
- Achieved 70% accuracy on the NTU-RGBD training dataset.
- Real-time performance averaged 15 frames per second (FPS) on the embedded system.
- Successfully demonstrated continuous human monitoring by the mobile robot.
Conclusions:
- The proposed method offers a viable solution for real-time human action recognition on mobile robots using RGB cameras.
- The system is computationally efficient enough for embedded platforms.
- This approach enhances the capabilities of robotic systems interacting with humans.
