Self-Powered Gesture Biometric Authentication via Triboelectric Dynamic Signature and Lightweight Deep Learning

With escalating security demands for smart devices, this study proposes a revolutionary self-powered gesture authentication system based on a triboelectric nanogenerator (TENG). Unlike conventional pattern-matching approaches, our flexible gesture unlock input panel (GUIP) captures multidimensional dynamic biometrics (e.g., pressure distribution, sliding velocity, and trajectory curvature) through a single-electrode TENG structure, enabling dual-layer identity verification. The GUIP exhibits exceptional stability (<±3% signal drift after 5,000 cycles) and flexibility (<5% fluctuation at 5 mm bending radius). Integrated with the lightweight MobileNetV4 architecture (model size: 2.6 MB), the system achieves 99.42% accuracy for recognizing five complex gestures in nine-grid unlocking scenarios and 99.27% user authentication accuracy by distinguishing individualized biometric features and maintains 0.2% false acceptance rate against spoofing attacks─representing an order-of-magnitude improvement over traditional methods. This work pioneers TENG-driven dynamic biometric encryption, offering a power-free, high-security solution for smart locks, mobile devices, and access control systems.