Integration of FEM-Guided Wearable Sensing and AI for Adaptive Upper Limb Rehabilitation Monitoring

Upper limb rehabilitation is critical for neurological recovery. This paper introduces a sensor-based monitoring framework enhanced by finite element method (FEM) simulation and artificial intelligence (AI) integration for personalised rehabilitation. A COMSOL-based FEM model simulates hand-object interaction to guide the design and placement of sensors within a custom wearable glove. The embedded system acquires biomechanical data at 1 0 0 ∼ H z, filters the signals via Kalman estimation, and transmits them wirelessly to an AI-based classification engine. A Random Forest model achieved an AUC of 0.94 and a correlation of r = 0.84 with clinical Fugl-Meyer scores from 12 subjects, showing that the system is clinically relevant. This integration of physics-informed modelling and wearable sensing provides a promising pathway for remote, adaptive neurorehabilitation.