FFT only frequency analysis

Hi Mike,

I have newly started exploring quantitative EEG analysis and came across your excellent series of books all of which I have been reading. I would like to enquire from anyone here if the assumption of non-stationarity of a straightforward FFT gives ‘inaccurate results’ or is it that it just cant provide time-decomposition of the frequencies. Lets say I have a 5 second epoch with a variation of frequencies across its time-domain. Is the frequency spectrum via a straightforward FFT still reliable enough for peak frequency or peak amplitude extraction? I hope my question is clear. I am looking at abnormal epilepsy subject segments and their complex discharges.


Hi Mubeen. It depends on how strong the non-stationaries are over the time window. For a 5-second window, I think one FFT should be fine. You could also try using Welch’s method using 1-second or 2-second windows. But usually, this is more of a concern for task-based datasets, or for longer recordings on the order of minutes.

Hi Mike, thanks for your input and advise.