ИСТИНА

We propose a general-purpose algorithm for exploring frequency patterns and organizing them into separate commands. It uses statistical parameters of EEG spectral components to define brain patterns and create a probabilistic (Bayesian) classifier to recognize them. We have also designed a training paradigm that allows user to form reproducible commands and to perform adaptive tuning of machine throughout learning process. Our approach was successfully applied in 2-command (classes) interface and potentially could be used in those with multiclass problems.