Design expanded BCI with improved efficiency for VR-embedded neurorehabilitation systems

A general brain computer interface (BCI) usually consists of three main units known as preprocessing unit, feature selection unit and classification unit. In this paper, an EEG-based BCI with expanded structure is introduced that provides opportunity to improve efficiency of virtual reality (VR) embedded neurorehabilitation systems. The proposed BCI has to detect three different neuro-stimulations during specified motor imagery tasks and generate proper virtual neuro-stimulations for the avatar to do the task in the VR world. In the proposed BCI, discrete wavelet transformation (DWT) and multilayer perceptron (MLP) neural network are applied for preprocessing and classification, respectively; and an expounder is added to eliminate misclassifications which lead to wrong virtual neuro-stimulations. Offline EEG signals are applied to examine the proposed BCI and results are demonstrated.