A smart insole system with hierarchical nanofiber sensors for multimodal gait analysis and identification

Traditional biometric technologies, including fingerprint, iris, and facial recognition, have been widely deployed for identity verification but remain constrained by privacy risks and environmental sensitivities. In contrast, gait recognition has emerged as a promising behavioral biometric due to its non-intrusive nature and difficulty of being disguised. Herein, we proposed a gait recognition method based on multi-sensor fusion and multi-scale feature extraction. Plantar pressure distribution and limb acceleration data were synchronously acquired via a custom-developed smart pressure insole integrated with an inertial measurement unit (IMU), enabling an end-to-end recognition pipeline. A lightweight parallel dual-branch (LPDB) attention module was designed to reduce computational overhead, while a multi-scale feature fusion (MSFF) method effectively integrated heterogeneous sensor data, capturing structural invariants within gait cycles and enhancing cross-cycle consistency. The system achieved a recognition accuracy of 98.5%. Furthermore, the piezoresistive sensor fabricated for the smart insole incorporated disordered nanofibers to form a hierarchical micro-protrusion conductive network, yielding a high sensitivity of 45.4 kPa⁻¹ in low-pressure regimes, significantly surpassing conventional porous material-based sensors. In addition, the as-prepared sensor exhibited superior anti-counterfeiting properties to facial features or fingerprints. This work offered a robust, privacy-preserving identification solution with potential applications in medical rehabilitation, sports science, robotic control, and the artificial intelligence of things (AIoT).