Machine‐Learning‐Assisted Piezoresistive Pressure Sensors for Gait Monitoring and Classification

ABSTRACT

Flexible pressure sensors with high sensitivity are essential for accurately monitoring foot biomechanics in gait analysis. Herein, a flexible piezoresistive sensor is fabricated via a facile dipping‐reduction process, in which reduced graphene oxide (rGO) and carbon black (CB) are sequentially coated onto a porous polyurethane sponge to construct a 3D conductive network with a gradient pore architecture. The prepared rGO/CB/PU sponge sensor exhibits exceptional comprehensive sensing performance, featuring an upper detection limit of 600 kPa, a maximum sensitivity of 16.1 kPa ⁻¹ within the 500–600 kPa range and stable signal output over 3000 loading‐unloading cycles. When integrated into a smart insole comprising eight spatially distributed sensor units, the smart insole enables real‐time mapping of plantar pressure distribution across key anatomical regions during walking and running. By coupling the multi‐channel sensor data with a convolutional neural network, the system achieves highly accurate classification of foot postures (eversion, normal, and inversion) with a recognition accuracy of 99.23%, demonstrating robust feature extraction and reliable differentiation of complex gait patterns. This work presents a promising strategy for developing high‐performance wearable sensing systems with significant potential in personalized rehabilitation and intelligent gait disorder diagnosis.