Inquiry on EEG Signal Preprocessing for Motor Imagery

[sprog1]

Thank you for this excellent work, which offers a new perspective on the interpretability of EEG signals and has inspired my own work. I primarily focus on emotion recognition and am not very familiar with motor imagery. Therefore, I would like to inquire about the details of the EEG signal preprocessing.

In the KAN_MLP_BCI.ipynb file, the dimensions of X and Y in the "two class raw" section are [210120,], which I understand as treating each sample point as an independent sample. However, does this mean that each time point necessarily triggers motor imagery? In the "two class CSP" section, it follows the conventional method where an entire time segment is considered as an independent sample, with X and Y dimensions being [120,]. I would like to know if treating each sample point as an independent sample is a common practice in motor imagery, as I have rarely seen this approach in emotion recognition-related papers. Additionally, what accounts for the differences in preprocessing between the raw data and CSP data?

Thank you for your time and assistance.