Direct Ink Writing–Based Fabrication and Evaluation of Customized High-Density Surface Electromyography Electrode Arrays for Hand Gesture Classification

High-density surface electromyography (HD-sEMG) is increasingly important for wearable hand-gesture recognition and human–machine interface (HMI) applications. However, commercial wet electrodes still require conductive creams and adhesive foams, which can cause impedance instability, skin irritation, and hygiene issues during long-term use, limiting wearable practicality. Here, we developed a gel-free 36-channel stretchable HD-sEMG electrode using a Direct Ink Writing (DIW) process and validated its printing precision, electrical/mechanical stability, and machine-learning-based gesture classification performance. Ag interconnects and dome-shaped Ag/AgCl electrodes were printed on a TPU substrate and encapsulated with PDMS. Process optimization achieved a line width of 217 ± 9.01 µm and a dome height of 569.3 ± 19.9 µm at a 4 mm pitch, close to the target dimensions. Based on area-normalized impedance (|Z|·A), the printed electrode showed lower impedance than a commercial electrode across the measured frequency range. During a 6-hour wear test, repeated EIS measurements showed that the 1 kHz impedance stabilized from 1.66 MΩ to 0.95–1.02 MΩ, while the coefficient of variation remained around 1% at most time points. Under 30% strain for 1000 tensile cycles, the electrode maintained electrical continuity and mechanical stability. In forearm HD-sEMG experiments, the electrode achieved SNRs of 17.06–21.34 dB (30% MVC) and 26.07–26.26 dB (100% MVC), and CNN and CNN–ViT models classified eight hand gestures with accuracies of 91.1% and 92.2%, respectively. These results demonstrate that the DIW-based gel-free electrodes can reliably acquire multi-channel muscle signals and suggest that muscle-informed channel grouping can be utilized in machine-learning frameworks for gesture applications.