Wireless real-time monitoring based on triboelectric nanogenerator with artificial intelligence

Abstract

A RepNet-based wireless self-powered sensor system is designed by just two components with deep learning algorithm, which has simple structure and high accuracy even without integrated circuit. Triboelectric nanogenerator (TENG) directly power the artificial intelligence sensor, and the algorithm extracts and encodes the convolutional features and local temporal information from a video. To test this model, we assemble a test dataset of 192 videos, comprising 32 frequencies of TENG. We then show the real-time detection backend based on the RepNet. This deep-learning-based backend also works well and demonstrates great feasibility and potential in the applications such as counting the number of LED flashing, estimating the possibility of LED flashing and detecting the changes of frequency. It is a potential and novel approach for sensing and transmited information of TENG-based self-powered sensors.