0
$\begingroup$

I have an EEG labeled data, which is the data that used for training, And I want to segment those data based on the time of EEG signal (Time-based Epoching) as a preprocessing step, based on the nature of EEG signals which is best:

  1. To segment the data based on time domain, for example to take each 64 samples together.

  2. Using the windowing functions like gaussian and hamming window. also if the best was this one,is it better to make the windows overlapped or not?

$\endgroup$
  • 2
    $\begingroup$ Can you add more details to your question? What do you mean by EEG labeled data and how do you want to segment it? Labeled according to EEG channels, Event Related Potentials, epochs or something else? $\endgroup$ – Naveen Apr 2 '15 at 2:43
  • $\begingroup$ @Naveen, see the edit, if it's not clear please tell me to add more information. $\endgroup$ – user2162652 Apr 3 '15 at 9:19
0
$\begingroup$

Overlapping segments don't cause any problem, unless your experimental design convinces you that such segments should not be allowed

$\endgroup$
0
$\begingroup$

I assume you want to segment your EEG-data in order to use the single segments as labelled training inputs to some machine-learning algorithm. Here the labels usually each hold a reference to a start- and end-index of the signal. In such a case the indexes determine your EEG-segments directly. (I suppose that is what you mean with your option 1.)

Before segmenting you could of course extract some features along the signal that can then be segmented accordingly as well.

If you were to use a window function during segmentation, how exactly would you do it? A window function can be used in cases where you want to re-weight the signal for some specific purpose for example. It doesn't make sense here as far as I can tell.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.