User Tutorial:P300 GUI for Offline Analysis Tutorial

From BCI2000 Wiki
Revision as of 22:26, 9 January 2009 by Sbriskin (talk | contribs)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to navigation Jump to search
The printable version is no longer supported and may have rendering errors. Please update your browser bookmarks and please use the default browser print function instead.
  • Run src/contrib./Tools/P300GUI/P300_GUI.m
  • Close the Editor window that appears, and type in P300_GUI
  • In the upper left-hand column, click on ‘Select Training Data’
  • Select the data files collected from the runs for THE, QUICK, BROWN, and FOX
  • In the lower left-hand column, press the Generate Feature Weights button
  • A pair of graphs should be displayed, similar to those below.
    • The important feature is to notice that the red line in the upper graph, the accuracy, increases with more flashes
  • Close the window of the graphs
  • In the lower-middle column, type Training for '‘File Prefix:'’, and 1 for '‘File #s to Save’'
  • Click on '*.prm (v2)’
    • This will generate a .prm (parameter) file in the same folder as the data files
  • Returning to BCI2000, click on ‘Config’ and ‘Load Parameters’, selecting the new .prm file
  • In the ‘Application’ tab, change TextToSpell to JUMPS, and click ‘Set Config’
  • Press ‘Start’ to collect one more run of data
  • Once the last run of data has finished, go back to the P300_GUI program and click ‘Add Test Data’ in the upper right-hand column
  • Select the newest set of data, and click ‘Apply Feature Weights’ in the lower right-hand column - A new window of graphs will appear
  • The important thing to see from these graphs is where the red line in the upper graph reaches 100%.
    • In the example below, this is from 9 to 11, and then drops. This means that 10 flashes gives the best accuracy.
9to11.PNG
  • If this does not give an acceptable level of accuracy, (typically 95% or higher), repeat more test runs to collect more data files. Add these data files as above, and click Apply Feature Weights in the lower right-hand column to look for how many flashes is needed. Repeat as necessary.
  • In the lower middle column, type Trained for ‘'File Prefix’' and 1 for '‘File #s to Save’'
  • Click ‘*.prm (v2)’ again to generate a Trained parameter file
  • Close BCI2000, MatLab, and the P300_GUI programs
  • Open BCI2000 with the batch\P3Speller_<Your_Amplifier>.bat file
  • Click on ‘Config’, and ‘Load Parameters’
  • Load the base parameter file for your amplifier, and then the Trained parameter file.
  • In the ‘Filtering’ tab, set EpochsToAverage to the ideal number found by now many flashes were needed to attain maximum accuracy. In our example, this would be 10.
  • In the ‘Application’ tab:
    • Set NumberOfSequences to this number as well
    • Delete the contents of the '‘Text to Spell'’ field
    • Set '‘InterpretMode’' to online free mode
    • Make sure the '‘DisplayResults’' box is checked
  • Click ‘Save Parameters’ and name as you see fit. Typically this parameter file would be P300_<Your Amplifier>_<Volunteer's Intials>_Free_Spelling.prm This file is now the file used for that volunteer in later P300 Spelling sessions


To continue onto performing P300 spelling experiments, continue to Performing a P300 Spelling Session.