PSD (Welch) on ERP

Hello!

I'm a new user of Brainstorm.
I am currently trying to compute de PSD (Welch) on an ERP, to observe the effects the averaging on the frequencies of the signal (i.e does it delete the 50Hz power line noise, etc...).

However, I am not familiar with the algorithm, and despite the tutorial information, I cannot find a proper window length for the computation. My ERP length is 350 ms (-50 ms to 300 ms).

Would it be possible to get some additional information about how to chose this window length depending on the data of which you want the PSD?

Thank you in advance.

François

Hi,

Welcome as a new brainstorm user :slight_smile:

Doing a frequency analysis on such a short time period (350 ms) is quite tricky, you will only be able to capture the higher frequency content properly (>10 hz more or less). Especially when using welch's method which splits the signal up into several parts with overlap to compute an average periodogram.

In this case it might be better to do a simple fft, maybe with a window to reduce some spectral leakage, on the ERP.

Kind regards,
Steven

1 Like

Hi Steven,

thank you for your quick reply.
Since I am interested in the 10Hz frequency, I guess performing a frequency analysis on ERP might not be a good idea.

Thanks again!

François

Indeed, computing an FFT on an ERP is never indicated and will not be very informative.

You could compute the average of the FFT power over the all your trials. For this, put all the epochs in the Process1 box and run "Frequency> Fourier Transform (FFT)" with the option "Save average power" selected. It would give you a measure a bit similar to the PSD over the continuous signals.

2 Likes

Okay, thank you for the tip!

François