Data Preprocessing

22/11/17

1. What has been done:

   1. Studied mne I/O, learned about EEG data visualization and analyzed different peculiarities of given data.
   2. After analysis of given data build different hypothesis and tasks for the further research.

2. Some ideas for further EEG channels processing:

   • Look on the channels covariances. Depending on the covariance values reduce uninformative channels. There are two ways of calculating the covariance:

     • Compare EEG channels on the whole time period
     • Divide the whole time period into time intervals and calculate covariances on them independently. Some difficulties with this approach: validity of the approach and ambiguity of the division (in EEG Emotions Recognition: the inability to distinguish proper time interval for the emotions)

   • Group some channels together by:

     • Channels covariance
     • Knowledge in neurobiology

   • Splitting general trends and local fluctuations in channels

   • Transform all non-stationary channels

   • Highlighting "seasoning"

   • Learn how to detect drop out. There might me problems with (Example, FP2 in resting_state/zavrin_open_eyes_eeg_15021500.vhdr, 20-25 seconds).

3. Tasks and solution ways:

   1. Implement channels correlation depending on validity of time and channel groups division
   2. There is no info about channels location: restore the localization if necessary.