Deep learning-enabled self-powered bimodal flexible sensor for intelligent access control

Area of Science:

• Materials Science
• Electrical Engineering
• Computer Science

Background:

• Intelligent access control systems are vital for smart architecture and urban security.
• Traditional biometrics face challenges like power dependency, privacy risks, and environmental sensitivity.
• A need exists for low-power, high-security, and multidimensional sensing solutions.

Purpose of the Study:

• To develop a self-powered bimodal sensor for intelligent access control.
• To provide a low-power, high-security, and multidimensional sensing solution.
• To address the limitations of existing biometric technologies.

Main Methods:

• A single-electrode triboelectric nanogenerator with a polydimethylsiloxane triboelectric layer and micro-pyramid array was designed.
• Contact electrification and electrostatic induction were used to convert mechanical stimuli into electrical signals.
• A deep learning framework utilizing a convolutional neural network was implemented for signal analysis.

Main Results:

• The sensor achieved high recognition accuracies: 99.83% for material identification and 98.88% for user authentication in single-dimensional tasks.
• The system maintained a recognition accuracy of 96.15% even in challenging environments.
• The device effectively extracts material electronegativity and human kinetic information.

Conclusions:

• The developed self-powered bimodal sensor offers a robust technological foundation for future smart security.
• This technology has potential applications in flexible electronic skins and personalized healthcare monitoring.
• The study demonstrates a novel approach to human-machine interaction in security systems.