Deep learning-assisted piezoresistive pressure sensors with broad-range ultrasensitivity for wearable motion monitoring

Abstract

Flexible mechanical sensors hold significant promise for motion assessment and biomechanical analysis. However, achieving high sensitivity and a wide operating range at low cost remains a major hurdle in pressure sensors. Herein, we propose a novel strategy for constructing a conformal force-sensitive interface on textile fiber structures. Specifically, multi-walled carbon nanotubes (MWCNTs) are deposited onto highly compressible polyester-based velcro textiles (PVT) via a low-cost spraying process. Benefiting from the strong synergy between the spraying technique and the PVT fiber structure, the intrinsic microstructure of PVT is preserved while forming highly interconnected conductive pathways, significantly enhancing the piezoresistive performance. Hence, the as-fabricated sensor demonstrates exceptional sensitivity of 3656.8 kPa⁻¹ (0–100 kPa) and an ultrawide detection range (0–3000 kPa), allowing for precise measurement of subtle pressures generated by breathing and high pressures exerted on human feet. Leveraging the scalability of this fabrication method, we develop a 16 × 16-pixel sensor array for spatial pressure mapping. Additionally, we design a multi-channel sensing insole system that, with the assistance of deep learning, accurately estimates vertical ground reaction forces (vGRF) across varying gait speeds, achieving an accuracy exceeding 98 %. More importantly, it enables continuous monitoring of vGRF variations during outdoor runs on different terrains. This work presents an affordable and scalable method for manufacturing flexible pressure sensors with high sensitivity and broad range, paving the way for applications in health monitoring, sports performance evaluation, and rehabilitation care.