PSD, welch method

Hi, I am new to Brainstorm and was wondering in-order to analyse the PSD for ~1sec pre-stimulus EEG activity, what the ideal window length would be? Also, If I should standardize the spectrum to view results (Freq bands 1-60 Hz) accurately?


Hi, 1 sec pre-stimulus EEG is too short, why not record more EEG data, I suggest at least 10 mins

Wang Yong

I guess what you want to do is an Event Related analysis in the frequency, if that is the case I suggest you to use measures like: Wavelet or Event Related Spectral Perturbation: That allows you to get information about the spectral changes in short times epochs like the pre and post stim interval. The Wavelet Analysis can be done in Brain Storm Software.

Welcome to BS community by the way :grinning:

thanks for suggestion, and how to join BS community?


You already did it! ,:grin:

If you have many trials with a baseline of 1 second for each trial, you can compute the FFT of each 1s block and average them (something similar to what the PSD process does on continuous signals). You can do this by dropping all your imported trials in the Process1 box, and run the process "Frequency > Fourier transform (FFT)" or "Frequency > Power spectrum density (Welch)": time window [-1s,0s], estimator > 1s, save average.
Both processes should give you the same results.

Standardize in what way?
To represent better the different frequency bands, you can use log-scale for either/and X and Y axis. You could also use the process "Extract > Extract values" to keep only a specific frequency band of interest.

Thank you for the helpful comments