bci2000.org BBS

can rythm isolation ( mu and beta ) be done on EEG.data before running ICA and the resulting activation matrix ( obtained directly as a matrix in eeglab ) be used for feature extraction ? However , the following paper https://arxiv.org/ftp/arxiv/papers/1312/1312.2877.pdf has done it differently . It performed ICA first and then performed rythm isolation which changes only the EEG.data ( say EEG.data_new) and then asks to perform weights*sphere*(EEG.data_new-meandata) . Which one is correct ? Thank you .

Samhitha,

if you bandpass filter your data into the relevant physiological bands (e.g., theta, alpha, beta, low gamma, high gamme, etc) it will be easier to make sense of your ICA components. Also remember that these physiological components have different amplitudes and SNR values, as the EEG signal follow the 1/f law, which is another reason to bandpass filter.

Regards, Peter

Thank you .