Self-Powered Tactile Sensor for Gesture Recognition Using Deep Learning Algorithms

Abstract

A multifunctional wearable tactile sensor assisted by deep learning algorithms is developed, which can realize the functions of gesture recognition and interaction. This tactile sensor is the fusion of a triboelectric nanogenerator and piezoelectric nanogenerator to construct a hybrid self-powered sensor with a higher power density and sensibility. The power generation performance is characterized with an open-circuit voltage V_OC of 200 V, a short-circuit current I_SC of 8 μA, and a power density of 0.35 mW cm^-2 under a matching load. It also has an excellent sensibility, including a response time of 5 ms, a signal-to-noise ratio of 22.5 dB, and a pressure resolution of 1% (1-10 kPa). The sensor is successfully integrated on a glove to collect the electrical signal output generated by the gesture. Using deep learning algorithms, the functions of gesture recognition and control can be realized in real time. The combination of tactile sensor and deep learning algorithms provides ideas and guidance for its applications in the field of artificial intelligence, such as human-computer interaction, signal monitoring, and smart sensing.