Machine learning-assisted wearable triboelectric-electromagnetic vibration sensor for monitoring human rehabilitation training

Wearable vibration sensors for human motion monitoring are beneficial for doctors to observe the process of patient rehabilitation diagnosis and treatment in real-time. However, realizing smart sensing and self-powering of wearable devices is still challenging. Herein we developed a portable and reproducible triboelectric-electromagnetic vibration sensor (TE-VS) composed of a rolling-type triboelectric nanogenerator (R-TENG) and a slider-type electromagnetic generator (S-EMG), which aims to achieve highly sensitive detection and rapid identification of human motion. The ingenious mechanical design of the TE-VS enables the R-TENG component to adopt tangential contact friction and the S-EMG component to mutually move between magnets and coils in a non-contact way, which greatly improves the capture efficiency of weak motion, reduces material wear and increases the service life of the device. Multi-channel self-powered sensing signals fusion in a miniaturized space aims to enhance the algorithms detection accuracy of TE-VS. Importantly, the spring damping system improves the device's response speed to external vibrations and reduces the impact of shocks, which contributes to improving operation stability. The output responses of different frequency responses and different accelerations were characterized, which validated the potential of TE-VS in gesture recognition and limb disorder detection. Additionally, TE-VS was also demonstrated as a powerful vibration energy harvester to fully charge a μF-level commercial capacitor in a short time. The proposed TE-VS realizes battery-free operation, solves the long-term power supply problem of wearable devices, and facilitates the real-time monitoring application of human rehabilitation exercises. This is definitely a promising application for human rehabilitation medical devices.