Gesture recognition systems epitomize a modern and intelligent approach to rehabilitative training, finding utility in assisted driving, sign language comprehension, and machine control. However, wearable devices that can monitor and motivate physically rehabilitated people in real time remain little studied. Here, we present an innovative gesture recognition system that integrates hydrogel strain sensors with machine learning to facilitate finger rehabilitation training. PSTG (PAM/SA/TG) hydrogels are constructed by thermal polymerization of acrylamide (AM), sodium alginate (SA), and tannic acid-reduced graphene oxide (TA-rGO, TG), with AM polymerizing into polyacrylamide (PAM). The surface of TG has abundant functional groups that can establish multiple hydrogen bonds with PAM and SA chains to endow the hydrogel with high stretchability and mechanical stability. Our strain sensor boasts impressive sensitivity (Gauge factor = 6.13), a fast response time (40.5 ms), and high linearity (R2 = 0.999), making it an effective tool for monitoring human joint movements and pronunciation. Leveraging machine learning techniques, our gesture recognition system accurately discerns nine distinct types of gestures with a recognition accuracy of 100%. Our research drives wearable advancements, elevating the landscape of patient rehabilitation and augmenting gesture recognition systems' healthcare applications.
Keywords: gesture recognition; hydrogel; machine learning; rehabilitation training; strain sensor.