This study aims to investigate whether combining "motor imagery" (the mental visualization of movement) with robotic exoskeleton therapy can improve upper limb recovery in stroke patients. Conventional robotic therapy often involves passive movement driven by the machine. In this study, patients in the experimental group will use their own voice to trigger the robot (e.g., saying "bend arm") while simultaneously imagining the movement. The research will compare this voice-enhanced "active" approach with traditional passive robotic training to see if it better promotes brain-to-muscle signal recovery and improves overall arm function .

Background and Rationale: Stroke is a leading cause of adult disability, and the "golden period" for rehabilitation requires intensive therapy to trigger muscle movement. Recent theories suggest that combining motor imagery with physical activity-known as Brain-Computer Interface (BCI) concepts-may be more effective than passive movement alone. However, complex EEG-based BCI is difficult to implement clinically. This study proposes a simplified "voice-controlled" interface to simulate the motor intent-to-action loop.

Study Design: This is a randomized, assessor-blinded, parallel-group clinical trial involving 32 subacute or chronic stroke patients. Participants will be randomly assigned to one of two groups for a 4-week intervention (5 sessions per week, 30 minutes per session).

Interventions:

Experimental Group (Voice + Exoskeleton): Patients will use the Nimbo single-joint robot. Before each movement, they will focus on imagining the target action for 2 seconds, then issue a standardized voice command (e.g., "bend hand") to trigger the robotic assistance.

Control Group (Passive Exoskeleton): Patients will receive conventional robotic-assisted training where the Nimbo exoskeleton moves the limb through a preset path without voice prompts or intentional imagery requirements.

Outcome Measures: Evaluations will be conducted at baseline (T0), week 2 (T1), and week 4 (T2).

Primary Outcome: Upper limb motor function measured by the Fugl-Meyer Assessment - Upper Extremity (FMA-UE).

Secondary Outcomes: Brunnstrom stage (recovery phase), KVIQ-10 (subjective imagery), Laterality Judgment Task (objective imagery processing), Sense of Agency (SoA) questionnaire, and System Usability Scale (SUS).

Safety and Monitoring: Safety checks, including pain (NRS), fatigue (RPE), and skin inspections, will be performed after every session. Any adverse events will be reported to the Institutional Review Board (IRB) and data safety monitoring personnel .

Inclusion Criteria:

• Ischemic or hemorrhagic stroke diagnosed by CT or MRI
• Stroke onset occurred at least 1 month prior to enrollment
• Aged 20 to 80 years
• Moderate to severe upper limb impairment (Fugl-Meyer Assessment-Upper Extremity score between 13 and 47)
• Cognitive function sufficient to follow instructions (MMSE score > 20)
• Stable medical condition with no severe pain in the affected upper limb

Exclusion Criteria:

• Severe spasticity of the affected upper limb (Modified Ashworth Scale score > 2)
• History of other neurological diseases (e.g., Parkinson's disease, brain tumor)
• Serious orthopedic conditions or skin lesions on the affected upper limb that prevent robot use
• Unstable epilepsy or severe psychiatric disorders
• Pregnancy