Dynamic Decoding of Intention via EEG

We consider the problem of dynamically decoding human intention via electroencephalography (EEG). We present two hierarchical frameworks to approach this problem. One framework processes the activities recorded in the sensor space, i.e., scalp, directly, while the other considers these activities in the source space, i.e., cortex. In both frameworks, the source-informed segmentation algorithm is used to identify the time intervals during which the spatial distribution of active cortical functional networks remains quasi-stationary, and a recurrent neural network (RNN) is employed as the dynamic classifier. The frameworks are applied to experimental motor execution vs. imagery data, and the results are discussed.