Brain Robot Interface

Nota:

• 64 electrodes

1. Scan for bad electrodes

Nota:

• Multivariate Pattern Classification. The system for learning to classify samples of EEG data

1. Training

   Nota:

   • Training the classifier on samples of data to recognize new samples with unknown class
2. Testing

   Nota:

   • Presenting samples of unknown class to the MVPC and see the performance, i.e. the accuracy of the classifier.
3. Gaussian Units

   Nota:

   • neural network with gaussian prototypes
4. SVM

   Nota:

   • Support Vector Machine, with kernel trick
5. LDA

   Nota:

   • Fisher's Linear Discriminant Analysis 

1. Feature Selection

   Nota:

   • The Power Spectral Density is estimated within frequency bands and presented as features.
2. Spatial Filter

   Nota:

   • Surface Laplacian or fit to single dipole
3. Bandpass filter

   Nota:

   • 1-48Hz or narrower
   1. "only HP-filter continuous data"

      Nota:

      • comment from Mike x Cohan's book neural ....