Short description of the method used for classification: 1. Bandpass filtering of the data to [8 12] Hz. 2. Selection of several channels by evaluation of the performance of a fitted Gaussian Mixture Model on the training data. 3. Fitting a Gaussian Mixture Model (of two Gaussians) on the test data. 4. Identify the correspondence between class label and cluster number via the cluster mean in feature space (low mean ~ class -1 (as in the training data)).