Page 1 of 1

about Mario Mu Rhythm Off-line Analysis

Posted: 12 Jul 2010, 23:21
by normal
We are trying to train the Mu Rhythm motor imaginary task with 4 states (left/right hand, both feet, facial movement) with Offline Analysis. Although we have succeeded in training two states's task by manual classifier configuration, the correctness of such configuration is very low for 4 states. Are there any tools available for automatic calculation the optimal classifier weights as P300GUI for P300 spelling task? Any suggestions will be appreciated.

Posted: 13 Jul 2010, 09:13
by jawilson
Hello,
Unfortunately, there is currently no such tool for the sensorimotor rhythm tasks, although something may be in the works. Fortunately, with the motor tasks, you have a good a priori idea of which electrodes (C3, Cz, C4, plus T7/T8 for facial, I assume?) and frequency ranges (mu, beta) will be changing, so even if the Offline Analysis gives poor initial results, you can select similar values on these electrodes for the initial cursor runs, and re-run the analysis later on the CursorTask data, instead of the StimulusPresentation data.
Let me know if you need more info on this.
Adam

Posted: 14 Jul 2010, 01:53
by normal
Thank you Adam, for the advice.

This is what we have done. We will check whether the training will improve the performance.

Hommy
jawilson wrote:Hello,
Unfortunately, there is currently no such tool for the sensorimotor rhythm tasks, although something may be in the works. Fortunately, with the motor tasks, you have a good a priori idea of which electrodes (C3, Cz, C4, plus T7/T8 for facial, I assume?) and frequency ranges (mu, beta) will be changing, so even if the Offline Analysis gives poor initial results, you can select similar values on these electrodes for the initial cursor runs, and re-run the analysis later on the CursorTask data, instead of the StimulusPresentation data.
Let me know if you need more info on this.
Adam