A dual conductive network-constructed ionogel sensor for human facial expression and action recognition based on the BDAA/AgNWs composition

Though ionogels serve as a key component of wearable and intelligent sensing systems, they possess several characteristics that constrain their practical applications, including poor electro-mechanical synergy, insufficient interface synergy (resulting in low ion migration efficiency and conductive network fracture under large strains), scarce data labels, and inadequate long-term feature extraction capabilities in recognition systems. To address these issues, a new butanediol diacrylate/acrylic acid‑silver nanowire (BDAA/AgNWs) ionogel was synthesized using 365-nm ultraviolet (UV) polymerization, with 1-butyl-3-methylimidazolium chloride ([BMIM]Cl), 2-acrylamido-2-methylpropane sulfonic acid (AMPS), and acrylamide (AM) as the monomer-ionic liquid system, dimethyl sulfoxide (DMSO) as the solvent, and silver nanowires (AgNWs) as the filler. The ionogel formed a dual conductive network via metal-ionic liquid/monomer chemical coordination, had a conductivity of 1.394 S/m at 386.81% stretch, good stability, and biocompatibility. The outstanding mechanical properties of ionogel enable its use as a highly sensitive strain-resistive sensor. When mounted on the skin, mechanical deformations induced by facial expressions or joint movements stretch the internal conductive network, resulting in predictable and reproducible changes in electrical resistance. These resistance variations are captured via a voltage divider circuit and converted into digital signals through an analog-to-digital converter, providing a stable and reliable input for subsequent intelligent processing. To overcome existing accuracy limitations in motion recognition, an advanced algorithm integrating Transformer-based temporal feature extraction with triplet contrastive learning is employed. This end-to-end sensing-to-recognition framework enables the system to accurately interpret dynamic physiological signals, achieving a facial expression recognition accuracy of 97.62%—a 15% improvement over mainstream algorithms—and a 10% enhancement in action recognition performance. These results establish a robust foundation for the advancement of flexible, wearable intelligent sensing systems and next-generation human-machine interaction technologies.