The invention relates to a steady state visually evoked potential+motor imagery (SSVEP+MI) brain-computer interface-based stroke rehabilitation training system and method. The system comprises a computer (1), an electroencephalogram cap (2), a mechanical arm (3), augmented reality (AR) goggles (4), an electrical stimulation feedback instrument (5), a rehabilitation pneumatic hand (6), an SSVEP stimulation interface (7), an SSVEP online training unit (8), an MI interactive interface (9), an MI offline training unit (10) and an MI online training unit (11). The method comprises the following steps: in first-stage rehabilitation training, an SSVEP brain-computer interface is combined with AR and the mechanical arm, SSVEP-EEG is decoded through a filter bank canonical correlation analysis (FBCCA) algorithm, which can help a patient to express intention and assist living, and meanwhile, the attention and cognitive ability of the patient can be trained, thereby laying a foundation for second-stage rehabilitation training. In the second-stage rehabilitation training, the MI brain-computer interface is combined with the AR and rehabilitation peripherals, MI-EEG is decoded through the FBCSP algorithm, a closed-loop rehabilitation loop can be formed through multiple feedback modes, and functional control connection between external limbs and the brain is repaired to help the patient to recover a limb movement function.