Sensor-Based Hand and Upper-Limb Movement Intention Recognition in Post-Stroke Rehabilitation: A Scoping Review

Background: Sensor-based systems increasingly allow post-stroke rehabilitation devices to infer voluntary movement intention and respond to active effort, but approaches vary widely in sensor modality, target movement, recognition method, and application. This review mapped sensor-based hand and upper-limb intention recognition after stroke. Methods: Following the Population–Concept–Context framework and PRISMA-ScR, PubMed, Embase, CINAHL, MEDLINE, and PEDro were searched from inception to 21 July 2026, and IEEE Xplore was additionally searched on 24 September 2026. Studies using sensor-based systems to classify, decode, trigger, assess, or provide feedback on voluntary hand or upper-limb movement in adults with stroke were included and narratively synthesized. Results: Of 1576 records, 17 studies were eligible: 5 electromyography (EMG)-based or myoelectric systems, 7 electroencephalography (EEG)/brain–computer interface (BCI)-based systems, 2 EMG-triggered functional electrical stimulation (FES) systems, and 3 multimodal assessment or feedback systems. Target movements were mainly hand opening/closing, grasping, and finger, wrist, or reaching movements. Applications ranged from intention-related physiological assessment and movement/performance feedback to real-time closed-loop control of robots, orthoses, exoskeletons, and FES. True multimodal sensor fusion was uncommon. Conclusions: Sensor-based intention-responsive rehabilitation after stroke has developed mainly through parallel EMG-based and EEG/BCI-based approaches, with limited multimodal integration and no standardized framework. The present review distinguishes direct intention decoding, sensor-triggered control, movement/performance feedback, and intention-related physiological assessment; future work should apply these distinctions consistently and advance multimodal integration and clinical validation.