Porous PDMS/SrTiO3 Composite Triboelectric Nanogenerators toward Self-Powered Information Encoding and Intelligent Gesture Recognition

Abstract Rapid advances in wearable human–machine interaction (HMI) and internet of things (IoT) systems are severely limited by the short service life, poor mechanical conformability, and environmental drawbacks of conventional battery-powered sensing devices. Triboelectric nanogenerators (TENGs) offer a compelling self-powered sensing alternative, yet real-world deployment is hindered by insufficient surface charge density and inadequate mechanical compliance of prevailing polymer friction layers. Herein, we adopt a facile sacrificial salt template strategy to fabricate flexible, three-dimensionally interconnected porous polydimethylsiloxane/strontium titanate (PDMS/SrTiO3) composite films for simultaneously boosting triboelectric output performance and mechanical compliance. The interconnected porous architecture greatly enlarges the effective contact area and improves elastic deformability of the active layer, while uniformly dispersed high-dielectric SrTiO3 nanoparticles reinforce dielectric polarization and interfacial charge retention. The optimized dual-structured TENG (DS-TENG) delivers a peak open-circuit voltage (VOC) of 178 V and a maximum power density of 450 mW/m2, retaining stable electrical stability over 20,000 continuous contact-separation cycles. Leveraging the high signal fidelity of DS-TENG, we demonstrate a wearable self-powered Morse code communication system, which encodes all 26 Latin alphabet letters from finger flexion-derived electrical signals. Furthermore, a five-channel DS-TENG sensor array integrated with continuous wavelet transform (CWT) and squeeze-and-excitation convolutional neural network (SE-CNN) achieves 92% training accuracy and 87% validation accuracy for recognizing ten complex numeric hand gestures. This work establishes an integrated self-powered sensing framework that bridges material–interface engineering with deterministic information encoding and data-driven gesture recognition, providing a feasible route toward practical wearable human–machine interaction hardware.