3D human pose detection using nano sensor and multi-agent deep reinforcement learning.

Area of Science:

• Biomedical Engineering
• Artificial Intelligence
• Sensor Technology

Background:

• Standard sensors struggle with the complexity of 3D human pose, limiting the accuracy of pose detection.
• Subtle changes in human motion are difficult to capture, necessitating advanced sensing and analysis techniques.

Purpose of the Study:

• To develop a novel 3D human motion pose detection method.
• To enhance the accuracy and reliability of 3D human pose detection using innovative sensor and AI technologies.

Main Methods:

• Integration of Nano sensors to collect human electromyogram (EMG) signals from key body parts.
• Application of blind source separation for EMG signal denoising, followed by time-domain and frequency-domain feature extraction.
• Development of a multi-agent deep reinforcement learning model for pose detection based on extracted EMG features.

Main Results:

• The proposed method achieved high accuracy in detecting diverse human poses.
• Quantitative results demonstrated excellent performance: accuracy (0.97), precision (0.98), recall (0.95), and specificity (0.98).
• The system significantly outperformed existing methods in 3D human pose detection accuracy.

Conclusions:

• The novel combination of Nano sensors and multi-agent deep reinforcement learning offers a highly accurate solution for 3D human pose detection.
• The method's effectiveness and high precision make it suitable for widespread application in fields such as medicine, film, and sports analysis.