Dual-scale bionic sensor with high stretchability for AI-enabled squatting motion monitoring

Real-time monitoring of knee flexion during sports is critical for injury prevention, yet precise measurement remains challenging due to large flexion amplitudes and complex velocity-frequency characteristics. To address this challenge, we propose a bionic honeycomb-shaped flexible resistive strain sensor (HSFRSS) with dual-scale structures, where the gauge factor (GF) and working range are enhanced markedly. Dual-scale structures not only enhance the stretchability of the flexible substrate but also increase the adhesion area of multi-walled carbon nanotubes (MWCNTs) on the substrate, thereby maintaining the continuity of conductive pathways under tensile deformation. The HSFRSS exhibits a high GF (up to 541), a wide working range (0-200% strain) and excellent durability (no degradation after 1000 cycles). Artificial intelligence is adopted to identify the sophisticated flexion information. An optimized decision tree regressor precisely maps resistance changes to knee joint angles with 99.9% prediction accuracy, enabling real-time angle estimation and immediate feedback, while maintaining strong generalization capability across different populations (prediction accuracy exceeding 97.8%). In parallel, a three-layer fully-connected neural network (FNN) distinguishes standard squats from error patterns in speed, amplitude, and inter-repetition interval, achieving 100% classification accuracy across different subjects. These results collectively demonstrate the significant potential of the proposed system for intelligent motion monitoring and posture correction in sports training applications.