PSD Issue
I am asking for your help because we are stuck in our study, we cannot understand why our PSD assumes strange values and consequently the possibility of evaluating a classical engagment index (alpha, beta, theta). The index assumes very low values that do not go beyond 20 and also the components of the waves have very low values. The pipeline that we developed on python, attached, should follow the classic steps for this kind of search: bandpass filter, notch filter, epoching date duration of 4 with sliding of 0.2 and the Welch for the PSD but the results are bad. U can also see it from the images here. We use a configuration of 2 channels with reference in the middle, fronthead and ground on ear. What we are missing? Something related to amplitude, errors in the pipeline or?
Thank you for helping, hope somebody could answer cause I'm gonna hit my head against the wall xD
FFT
Engagement index
Comments
Hello Clod,
You do not mention what electrodes / electrolyte you are using. Are you getting expected amplitudes of the time series? Alpha generation at the forehead is usually low. It will be much larger in occipital / parietal areas.
How do your signals look in the OpenBCI_GUI, as opposed to your Python? Especially the GUI FFT and Band Power widgets. Any different? If so, then suspect your Python processing chain.
Also try the Focus Widget, set for 'relaxation', which mostly tracks alpha generation.
https://docs.openbci.com/Software/OpenBCISoftware/GUIWidgets/#focus-widget
William
We are using dry snap electrodes with athletes ( no time for gel or wet) and we can use only forehead. The signals looks good in the GUI while the band power shows strange spikes. I was also wondering how it was computed the numerical value of the relaxation index that always takes rather low values in the widget
The dry electrodes with velcro band are HIGHLY dependent on how much pressure is applied. You may not be using enough tension on the band. Additionally, if the athlete has been or will be sweating, that will likely also affect readings erratically, depending on the circumstances. And facial movement such as speaking, eye movement, jaw movement, etc. can also throw artifacts into the EEG.
Another excellent alternative are the small adhesive foam (snap) pre-gelled electrodes, commonly used for EEG, ECG, EMG. There are some of these in the Shop, but these are widely available on Amazon and other medical supply stores. If you buy in bulk quantities, 20, 50, etc. -- are quite affordable and give better signal quality on skin areas. No dependence on application pressure since there is electrolyte inside the package.
A commonly used EEG position is called the 'mastoid' location, which is free of hair. This is usually applied using a foam dot.
With such a reference location, near occipital regions, you may pick up more alpha.
Do the spikes slow down if the subject is being still? If so, suspect the earlier mentioned artifacts.
I believe the Brainflow metric is using some type of ML model, with significant alpha weight.
William