Graphene-printed sheet textiles for AI-driven infant sleep position recognition with a resistance tomography-inspired approach

This article presents a graphene-printed textile sensor system for infant sleep position recognition, leveraging an electrical resistance tomography (ERT)-inspired approach combined with AI-driven analysis. The system consists of a crib-sized mattress covered with a coated fitted sheet (CFS), created via screen printing of a graphene nanoplatelet (GNP)-based ink onto fabric. The coated fabric exhibited piezoresistive behavior, with gauge factors of 9.9 (in the wale direction) and 1.2 (in the course direction) at a 2.7% strain. Voltage data, acquired using an ERT-inspired protocol, were analyzed through machine learning (ML) models to identify sleep positions, including potentially hazardous ones associated with conditions, such as sudden infant death syndrome (SIDS). The system was calibrated through experiments with a wooden cube and a wooden dummy, supported by finite element method (FEM) simulations to replicate the fabric’s mechanical and electromechanical behavior, enabling synthetic training data generation. Final validation of the CFS involved testing with a lifelike infant mannequin under three conditions: direct contact with the sensorized sheet, the presence of a blanket interposed between the mannequin and the sheet, and induced electrode malfunctions. The classification framework included: 1) general posture recognition (supine, lateral, and face-down); 2) detailed classification (DC) of 30 specific postures, including hazardous ones; and 3) a hierarchical strategy to improve classification robustness and generalization. The system achieved 99.4% accuracy in direct contact and 94.8% with a blanket placed between the mannequin and the sheet, confirming its robustness for real-world monitoring. The proposed solution offers a nonintrusive, scalable, and privacy-preserving alternative to traditional monitoring methods.