High‐Performance Hybrid Triboelectric Nanogenerator Integrated With Machine Learning for Human Activity Recognition

Rapid advances in wearable electronics have increased demand for sustainable, self‐powered systems for real‐time human activity monitoring without an external power supply. Triboelectric nanogenerators (TENGs) address this need by harvesting mechanical energy and producing electrical signals for activity recognition. Here, we report a TENG‐based wearable sensing system integrating a hybrid Au–rGO/PDMS composite TENG into footwear for real‐time human activity recognition. Au nanoparticles were synthesized and simultaneously embedded on rGO nanosheets by a single‐step, in situ electron irradiation method. Characterization confirmed uniform Au nanoparticle distribution and strong Au–rGO interfacial contact, increasing charge‐trapping site density and the effective dielectric constant. Incorporating Au–rGO nanostructures in PDMS yields a flexible, high‐output TENG with enhanced performance from enlarged contact area, efficient interfacial charge transfer, dense charge trapping, and reduced charge recombination. The optimized device delivers a peak power density of 10.62 W/m 2 , charges a 33 µF capacitor in 210 s, and powers 104 light‐emitting diodes by tapping, achieving nearly fourfold higher output than pure PDMS. For activity recognition, voltage patterns from the footwear‐mounted TENG were processed into time‐ and amplitude‐domain features. Using machine learning, random forest classifier distinguished four activities, no movement, walking, running, and jumping with 90.4% accuracy (leave‐one‐subject‐out, three volunteers), confirming reliable recognition for Internet of Things and health monitoring.