Hydrogel epidermal electrodes enabling high-accuracy surface electromyography-based intelligent prosthetics

With over half a billion people affected by limb deficiencies or amputations worldwide, there is a critical need for advanced prosthetics to restore motor function. Surface electromyography (sEMG)-based prosthetics is especially attractive due to easy signal acquisition, while the application is hindered by unstable skin-electrode interfaces and insufficient decoding accuracy, especially for individuals with limited muscle mass. This study presents a material-algorithm co-designed platform integrating a hydrogel epidermal electrode and a machine-learning decoding framework for intelligent prosthetic control. The hydrogel electrode, consisting of a dynamic hyaluronic acid network crosslinked via reversible boronate ester bonds and reinforced with Ti₃C₂T_x MXene nanosheets, exhibits conformal skin contact, rapid electrical recovery within 100 ms, a bio-interface adhesion strength of 8.5 kPa on porcine skin, and a hybrid ionic-electronic conductivity of 0.145 S m^-1. These features facilitate effective skin-electrode coupling and stable sEMG acquisition, achieving an SNR of 12.59 dB. Coupled with a bidirectional long short-term memory (Bi-LSTM) neural network incorporating multi-dimensional feature extraction, attention mechanisms, and contrastive learning, this system achieves a high classification accuracy of 96.91% and enables robust control of intelligent prosthetics. Real-time robotic hand control demonstrated low latency (< 40 ms) and intuitive operation for a volunteer with upper-limb loss. This unique material-algorithm co-design strategy advances bioelectronic interfaces, enhancing human-machine interaction and prosthetic functionality.