ObjectiveTo investigate the clinical effects of the mirror neuron system (MNS)‐based training on upper extremity motor function and cognitive function in stroke patients.MethodsSixty stroke patients (time from stroke onset 3–9 months) with upper extremity paresis (Brunnstrom stage II–IV) and cognitive impairment (MoCA score ≥ 15) were enrolled in this study. Patients were randomly allocated into MNS treatment group (N = 30) and control group (N = 30). Both groups underwent regular training for upper extremity motor function and cognitive function, and the MNS group was trained with a therapeutic apparatus named mirror neuron system training (MNST) including different levels of action observation training (AOT). Training lasted 20 min/day, 5 days/week for 8 weeks. MoCA, reaction time, and Wisconsin Card Sorting Test (WCST) were assessed at baseline and 8 weeks after training. Furthermore, Fugl‐Meyer assessment (FMA) and Modified Barthel index (MBI) were adopted to evaluated upper extremity motor function and daily life ability.ResultsAfter 8 consecutive weeks’ training, both groups showed significant improvements on the upper extremity motor function, cognitive function, and daily life ability score after training (p < .05). The MNS group showed significantly improved upper extremity motor function and cognitive function (p < .05) compared with control group.ConclusionsCombining MNS‐based and conventional training can improve upper extremity motor function and cognitive function in stroke patients.