Electroencephalographic (EEG) control of cursor movement in three-dimensional scene based on Small-World neural network

Contraposing mental task recognition, a novel Small-World neural network (SWNN) algorithm is proposed for the EEG classification tasks trying to solving the target hitting problem with the low spatial dispersion degree. A BMIs experimental paradigm with a series of Cue Scenes which focus on three-dimensional (3D) object control is described. Bring in the Small-World neural network, which is transformed from the regular network by random rewiring according to the rewiring probability P and learned by back-propagation rule. In order to find the better method of network structure design, a set of parameters about network constructing are fixed multifariously. After repeated authentication, we find a way to balance classification effectiveness and time complexity of the proposed algorithm. For emphasizing the greatest distinction among the most outstanding features of different movement directions, CSP with one versus rest strategy is used to carry out spatial filtering. Finally, the feasibility of the algorithms is primarily indicated through offline analyzing of EEG data. The non-marked trials judged originally by the method contained in BCI2000 system retrieve their new marks which agree with the cues of that time on the whole, While the algorithm framework comprised of CSP (OVR) and SWNN is used.