In EEG Motor Imagery dataset BCI Competition III ( Data set IVa ‹motor imagery, small training sets),How can I train the samples with the two class(1-left,2-right).?
In the dataset ,Classes are mapped with the position or (at the instant sample)..For example, left class is mapped with 47428 sample..should i take the values upto 47428 in input or only 47428th sample corresponds to 118 electrodes???
Hello Thenmozhi,
Usually EEG datasets provide a marker for the start of an event based on the experimental protocol followed to record the data. For example, consider 2 classes (1-LH and 2-RH), if the sample number 47428 is the starting sample for the RH imagery trial, and if the RH imagery trial was 3 seconds long with a sampling rate of 1000 Hz, then you would have to consider 3000 samples of that trial from 47428 to 50428 as belonging to class 2. The final matrix for a single trial would then be 3000 (samples) x 118 (channels).
Here is a link to Regularized Common Spatial Pattern Toolbox (RCSP) developed by Dr. Fabien Lotte's group for the motor imagery datasets.
Code: https://drive.google.com/file/d/0B93wwoGtlGkmazNiaE9hanlRQk0/view
Paper: https://hal.inria.fr/file/index/docid/520754/filename/tbme10.pdf
Hope you find the MATLAB code helpful for your analysis.
- Badges
- Science topic