A Flexible Cognitive Built Wearable Platform Using Multimodal Therapeutic Sensors and Reinforcement Learning Technique for Upper-Limb Stroke Rehabilitation

Stroke is one of the most common long-term motor disabilities which requires a personalized rehabilitation strategy capable of delivering reliable therapeutic interventions with continuous assessment. This paper proposes the integration of the inertial measurement unit, force-sensing technology, and electromyography (EMG), along with an artificial intelligence (AI)-based, cognitive–adaptive control platform and intelligent framework within a closed loop for precision neurorehabilitation. This research proposes a methodology to continuously monitor patients’ motor functions and compensate for movements, dynamically optimize rehabilitation, and detect fatigue intensity in real time. This proposed architecture combines deep learning-based movement analysis with multimodal sensor fusion and reinforcement learning-based adaptive control. This is done alongside cloud-based facility analytics for long-term longitudinal assessment and clinic medical assistant therapy management, and a lightweight edge computing module for feature extraction and cognitive decision-making. A total of 36 post-stroke patients with upper-limb hemiplegia were used in an 8-week clinical study, in which Fugl–Meyer assessment (FMA) scores increased by an average of 87.9%, EMG muscle activation by 42.8%, and therapy efficiency improved by 31.6%. Therapy efficiency improved by 32.8% and 21.9% rehabilitation reduction when compared with traditional physiotherapy. The composite performance index has rapidly increased to 0.949, indicating overall rehabilitation efficiency, with repeated measures analysis of variance confirming a statistically significant trend. Thus, the home-based stroke recovery with substantial potential for research constrains the healthcare environment with smart adaptive and real-time patient-specific optimization, which is proposed in this AI framework as an interpretable clinical deployment solution.