In this study, we propose an effective training method for 2-class motor imagery tasks on brain computer interface (BCI) systems viable even for distracting environments. For non-invasive BCIs, it is difficult to capture event-related desynchronization (ERD) and event-related synchronization (ERS) signals through electroencephalogram (EEG) in places where it is difficult for subjects to concentrate. To improve concentration under a distracting environment, our proposed training method implemented a graphical interface as a source of visual feedback. The performance of the implemented training method is evaluated by comparing its results with those of a training method that does not support visual feedback. The experiments are held while a variety of noises are produced to simulate a distracting environment. The results of the experiment demonstrate the effectiveness of the proposed training method in distracting environments for 2-class motor imagery tasks.