Deep Learning Enabled Neck Motion Detection Using a Triboelectric Nanogenerator.

Area of Science:

• Biomedical Engineering
• Wearable Technology
• Sensor Technology

Background:

• Cervical health is closely linked to neck motion.
• Accurate neck motion detection is crucial for healthcare intelligence.
• Existing solutions lack wearability, flexibility, power efficiency, and affordability.

Purpose of the Study:

• To develop a practical, wearable, flexible, power-efficient, and low-cost neck motion detector.
• To utilize triboelectric sensors and deep learning for robust neck motion recognition.
• To achieve high accuracy in identifying various neck movement states.

Main Methods:

• Integration of four flexible and stretchable silicon rubber-based triboelectric sensors onto a neck collar.
• Utilizing distinct voltage signal patterns from sensors to represent different neck motion states.
• Incorporation of a carbon-doped silicon rubber layer for external electric field shielding.
• Development of a convolutional neural network-based deep learning model for motion classification.

Main Results:

• The developed sensor group generates unique voltage signals corresponding to different neck motions.
• The carbon-doped layer significantly enhances the robustness of motion identification.
• The deep learning model accurately recognized 11 classes of neck motion (8 bending, 2 twisting, 1 resting).
• Achieved an average recognition accuracy of 92.63% for neck motion detection.

Conclusions:

• The developed neck motion detector is wearable, flexible, and self-powered.
• The combination of triboelectric sensors and deep learning enables robust and accurate neck motion detection.
• This technology shows significant promise for applications in neck monitoring, rehabilitation, and assistive control systems.