Hello, everyone. I am using the FOOOF method to extract peaks (Gaussian function). I need four peaks, each for one bandwidth. The Brainstorm allows me to consider four peaks, but not each for a bandwidth (0.5-4, 4-7, 7-13, 13-28). How can I solve this?
FOOOF algorithm tries to model the PSD using peaks where they are needed to minimize the error between the PSD and the model. Forcing the place of the peaks to specific frequency bands seems counter intuitive to the purpose of FOOOF, e.g., in the case when there are not visible peaks in the PSD for those bands.
@Luc, any comment on this?
1 Like
@Raymundo.Cassani is spot-on here: FOOOF usually fits peaks where their detection criteria are met and fit error is subsequently minimized, with little interest in the specific frequency band (as long as fit criteria are maintained, otherwise they are removed).
If I understand correctly, did you want to extract the peaks fit in all channels on the basis of their frequency range? If so, you could use some of the data fields in the Options.FOOOF structure which save peaks and provide their parameters as well as the channel from which they originate. Without seeing your data, I would warn that some channels may have two peaks in a range and some may have zero (you can keep the highest amplitude peak if you encounter the former).
Forcing peaks to exist in certain frequency ranges is possible using emerging methods, but this approach should be used with caution for the reason previously mentioned.
Hope this helps!
1 Like
Thank you @Raymundo.Cassani and @Luc. You are right. It is more logical to allow FOOOF to consider peaks elswhere it recognize for better fitting, even a band has two peaks while another band has no peaks.
A post was split to a new topic: FOOOF model, and PSD units
@emokh1986, please make a new post of a new question.
This will help to other users to find relevant information.
Your new questions was moved to:
FOOOF model, and PSD units
1 Like