Triboelectric Bending Sensors for AI-Enabled Sign Language Recognition

Area of Science:

• Wearable Electronics
• Human-Machine Interfaces
• Sensors and Actuators

Background:

• Contemporary wearable sensors require external power, limiting their application in the Internet of Things.
• Self-powered sensors are crucial for advanced human-machine interaction and diverse wearable applications.

Purpose of the Study:

• To develop a smart wearable system for sign language recognition using self-powered triboelectric sensors.
• To integrate artificial intelligence (AI) for accurate signal pattern recognition in wearable electronics.

Main Methods:

• Developed a system with five arch-structured self-powered triboelectric sensors and a data acquisition unit.
• Employed a slider-crank mechanism for quantitative sensor performance assessment.
• Utilized a long short-term memory (LSTM) network for sign language signal pattern recognition after noise reduction.

Main Results:

• Achieved a sign language recognition accuracy of 96.15% using two trained LSTM models.
• Successfully reduced environmental noise and cross-talk among sensor channels using low-pass filters.
• Demonstrated quantitative performance assessment of the triboelectric sensors.

Conclusions:

• This work presents a novel integration of triboelectric sensors with AI for effective sign language recognition.
• The developed self-powered system offers a new application for triboelectric sensors in wearable electronics.
• The system overcomes power limitations of conventional wearable sensors, enabling broader deployment.