 This paper proposes a novel data augmentation technique for reducing calibration time in the rapid serial visual presentation, RSVP, classification task. The technique uses a GM-based approach to generate artificial EEG data from existing majority class data, which helps to reduce the calibration time required for the minority class data. The results demonstrate that the proposed technique can significantly improve the classification accuracy of the model and reduce the calibration time compared to other approaches. This article was authored by Meng Siu, Yuan Fangchen, Yijuan Wang, and others.