A A Neural Network Approach for EEG Classification in BCI

A Brain Computer Interface (BCI) is a new communication channel allows a person to control special computer applications like a computer cursor or robotic limb through the use of his/her thoughts. BCIs had become an active research area in the last decade. BCI research is based on recording and analyzing electroencephalographic (EEG) data and recognizing EEG patterns associated with various mental states. Supervised classification methods are employed to recognize these EEG activity patterns to learn the mapping between the EEG data and classes corresponding to mental tasks. Selecting an optimal frequency band and extracting a good set of features is still an open research problem. In this paper, it is proposed to investigate EEG signals, extract features of motor imagery in the frequency domain using Hilbert transform, to compute the maximum and minimum energies. For efficient classification, Principal Component Analysis is used for feature reduction. Classification is carried out using multilayer perceptron with different learning rate and Momentum.