Hi,

To visualize the power spectrum of my artifact and component corrected EEG data, would Welch periodogram be best?

My objective to isolate subject specific alpha and beta ranges. The generalized alpha range of 8-13 Hz and beta range of 13-30Hz, seems to dilute the ERD.

How do I plot the power in different frequency bands as opposed each frequency?

Yes, Welch’s method tends to be better than one FFT over the entire time window in the presence of non-stationarities (which there are in EEG data).

I’m not sure what you mean by “power in different frequency bands as opposed each frequency.” Do you mean to compute the average power within a frequency range? That’s a good idea. That would then give you, for example, a bar plot with alpha and beta in two sets of bars.

My objective was to identify the frequency bands of interest for each subject that shows the max change in ERD. I would like to define individualized frequency ranges based on ERD for each task.

Right now I compute ERD for 2 Hz frequency ranges between 8 and 30Hz and identify the frequencies with max ERD. I would have liked to look at the power in each frequency and then compute the ERD based on the frequencies that show the greatest power.

I see. You can also sum the time-frequency power plot data over some reasonable time window. The FFT, Welch’s method, and time-summed time-frequency power are all basically the same things with various amounts of smoothing and temporal localization.

Hi priyatvb and Mike,

Hope you don’t mind me jumping in your conversation, but could I ask something in relation to what Mike just mentioned? It might be a very basic question though - if so I apologise.

So, if the FFT is done with windowing and averaging, would it be the same as Welch’s method? Or is there any standard way of windowining (or anything else if I’m missing any) to the calculation to be called the Welch’s method?

I’ve been reading quite a few things on these and got a bit confused by so many different names and formulas, as well as different results, or figures…

Very best,

Saya

Yes, Welch’s method is basically windowed FFTs. There are several windowing functions, and they generally produce similar results.