I'm trying to analyze and extract some frequency domain features for a simple machine learning model from EEG signals. I partitioned my data to 30s epochs and used np.fft to get the frequency and the (power/amplitude?) and for each frequency I categorized it to bands of Alpha/Beta/Delta/Theta. Finally for each epoch I get group by the frequency band and get band which has the highest average. Can this be improved ? The reason being that I get the same band (Delta) for every single epoch.

  • $\begingroup$ Interesting. Please paste in here all the code you made for this, and all the plots you are getting, and and we could give you a good answer. $\endgroup$
    – Brethlosze
    Jul 29, 2022 at 19:22
  • $\begingroup$ 30 s seems quite long. How quickly do the patterns change for the type of brain activity you are studying ? $\endgroup$
    – Hilmar
    Jul 30, 2022 at 13:32
  • $\begingroup$ Adding to the previous comments, it is unclear what you are trying to achieve. Are these resting state data or data recorded during a certain behavior? Is the neuro-phsyiological pattern in the frequency-domain or time domain? Are you trying to distinguish between two states? In other words, the context is unclear, which makes it impossible to tell you if "this" can be improved. $\endgroup$ Aug 1, 2022 at 5:52


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.