A BCI system for stroke subjects will be designed in an attempt to train the motor-related areas in their brains responsible for hand open/close function. A robot hand will be used and controlled by the system in real time based on EEG pattern for identification of motion intention for hand movements in the motor cortex. A randomized controlled trial study design will be adopted. All subjects will be assessed by clinical assessments including FMA and neuroimaging assessments using multimodal MR techniques before and after the training to evaluate training effectiveness.

Inclusion Criteria:

1. Sufficient cognition to follow simple instructions, as well as understand the content and purpose of the experiment;
2. Subcortical ischemic lesion within the territory of the middle cerebral artery;
3. Have moderate to severe motor disability at the paretic upper limb (assessed by Fugl-Meter Assessment (FMA), Modified Ashworth Score of fingers (MAS), and Action Research Arm Test(ARAT)).
4. Hemiparesis resulting from a single unilateral lesion of the brain with onset at least 6 months before data collection.

Exclusion Criteria:

1. Severe hand spasticity, open hand wound or hand deformity;
2. Visual field deficits;
3. Aphasia, neglect, and apraxia,
4. Participation in any therapeutic treatment ("outside therapy") performed with the paralyzed arm during the planned study - including baseline and follow up.
5. History of alcohol, drug abuse or epilepsy,
6. Bilateral infracts,
7. Uncontrolled medical problems,
8. Serious cognitive deficits,
9. Other MRI contraindications