Self-Powered Tactile Sensor for Gesture Recognition Using Deep Learning Algorithms.

Area of Science:

• Materials Science
• Electrical Engineering
• Artificial Intelligence

Background:

• Wearable tactile sensors are crucial for human-computer interaction.
• Existing sensors often lack sufficient power density and sensitivity.
• Integration with advanced algorithms is needed for complex tasks like gesture recognition.

Purpose of the Study:

• To develop a multifunctional wearable tactile sensor.
• To enhance sensor performance through hybrid nanogenerator design.
• To enable real-time gesture recognition and interaction using deep learning.

Main Methods:

• Fabrication of a hybrid self-powered sensor by fusing triboelectric nanogenerator and piezoelectric nanogenerator.
• Characterization of power generation performance (open-circuit voltage, short-circuit current, power density).
• Evaluation of sensor sensitivity (response time, signal-to-noise ratio, pressure resolution).
• Integration of the sensor onto a glove and application of deep learning algorithms for gesture recognition.

Main Results:

• Achieved high power generation: 200 V open-circuit voltage, 8 μA short-circuit current, 0.35 mW cm^-2 power density.
• Demonstrated excellent sensitivity: 5 ms response time, 22.5 dB signal-to-noise ratio, 1% pressure resolution (1-10 kPa).
• Successfully implemented real-time gesture recognition and control using the sensor and deep learning.

Conclusions:

• The hybrid self-powered tactile sensor offers superior power density and sensitivity.
• Deep learning algorithms effectively enable real-time gesture recognition and control.
• This technology provides a foundation for advanced AI applications in human-computer interaction and smart sensing.