A Subject-Specific Time Window Selection Method for Motor Imagery BCI

Brain-computer interface based on motor imagery (MI) electroencephalogram is a promising technology for the future. However, the time latency during the MI period exhibits variability among the trials of different subjects, which can significantly affect the segmentation of each subject's trial using the fixed time window. To address this issue, the study proposes a subject-specific time window selection (TWS) method based on wavelet transform. The experiment results show that the proposed TWS achieves the highest accuracy across all three MI classification methods, with classification accuracies of 77.19%, 83.98%, and 85.33% on average respectively.