Highly durable machine-learned waterproof electronic glove based on low-cost thermal transfer printing for amphibious wearable applications

Gesture recording, modeling, and understanding based on a robust electronic glove (E-glove) are of great significance for efficient human-machine cooperation in harsh environments. However, such robust edge-intelligence interfaces remain challenging as existing E-gloves are limited in terms of integration, waterproofness, scalability, and interface stability between different components. Here, we report on the design, manufacturing, and application scenarios for a waterproof E-glove, which is of low cost, lightweight, and scalable for mass production, as well as environmental robustness, waterproofness, and washability. An improved neural network architecture is proposed to implement environment-adaptive learning and inference for hand gestures, which achieves an amphibious recognition accuracy of 100% in 26 categories by analyzing 2,600 hand gesture patterns. We demonstrate that the E-glove can be used for amphibious remote vehicle navigation via hand gestures, potentially opening the way for efficient human-human and human-machine cooperation in harsh environments.