Deep Learning‐Enabled Multifunctional Electronic Skin System for Human Activity Recognition and Sign Language Translation

ABSTRACT The integration of wearable technologies and intelligent algorithms is anticipated to enable synergistic platforms for human motion monitoring and enhanced human‐computer interaction. However, the realization of personalized and multi‐functional integration of electronic devices through simple fabrication processes remains challenging. Additionally, the current artificial intelligence algorithms lack the universality across different application scenarios. In this work, we present a multifunctional electronic skin (e‐skin) system with deep learning‐assisted multi‐sensing capabilities. The e‐skin system integrates a dual‐layer 3D‐printed e‐skin, a flexible multi‐channel wireless acquisition circuit, a robust data processing framework, and a multi‐action classification neural network (MAC‐Net) designed for human activity and sign language recognition. The e‐skin can monitor finger bending, joint strain, temperature, humidity, and glucose. Its integrated data processing framework enables an automated workflow from raw signal acquisition to model training and evaluation, enhancing processing efficiency and signal reliability. Experimental results demonstrate that the integration of e‐skin with MAC‐Net enables accurate daily motion monitoring, sign language recognition, and human‐computer interaction, offering a practical and customizable solution for next‐generation wearable systems.