A flexible pressure sensor based on a PVA/TA/CNT organogel for deep learning powered human gesture recognition

Abstract

Gel-based pressure sensors are widely used in wearable electronics to provide features of human activity detection like gesture, gait, and sitting posture recognition. Herein, we report a double-network (DN), non-swelling, and adhesive organogel by means of physical cross-linking. Two types of polyvinyl alcohol (PVA) were mixed at a mass ratio of 1 : 1.4 and used as the matrix of the hydrogel. 1.5 wt% of tannic acid (TA) was added as a cross-linking agent, and 2 wt% of carbon nanotubes (CNTs) was added as both the reinforcing framework and conductive material. After freeze–thaw treatment, the performance of the resistive pressure sensor was greatly improved. Fabricated using a binary solvent system of ethylene glycol (EG) and deionized (DI) water, the PVA/TA/CNT organogel exhibited excellent strength. The PVA/TA/CNT organogel sensor exhibited a sensitivity of 9.088% kPa⁻¹, the highest pressure detection of 130.3 kPa, a response/recovery time (230 ms/450 ms), and excellent repeatability (2000 cycles). Additionally, a square-shaped organogel pressure sensor with a 3 cm side length was prepared. Combined with a circuit containing a microcontroller, multiplexer, operational amplifier, and other components, 13-channel resistance data acquisition was realized. This corresponded to a total of 156 virtual tactile points on the sensor to react to pressure on different sensor components. Four widely used gesture recognition types were achieved with an accuracy of 98.75% using the Convolutional Neural Network-Multilayer Perceptron (CNN-MLP) model. This study demonstrates the enormous potential of organogel-based materials for the creation of intelligent, flexible pressure sensors with exceptional stability.