Effects of Different Feature Vectors and Neural Network Topology on EEG Mental Tasks Classification

K. Tavakolian, A.M. Nasrabadi, S.K. Setarehdan, M.A. Khalilzadeh, R. Miri, and M. Falaknaz (Iran)


Linear and Nonlinear features, Leave-One-Out-Method,feature selection algorithm, Genetic Algorithm


In this research some important aspects of mental task classification is investigated. The effect of neural network topology and structure on the accuracy of classifications is evaluated. One of the implemented approaches toward this problem is leave-one-out-method or LOOM which is more reliable than others. According to obtained results a neural network with one hidden layer of 20 neurons was proposed as the optimal structure for classification of mental tasks. The next phase of the research aims at giving an intelligent feature selection algorithm so an especial real coded genetic algorithm is implemented.

Important Links:

Go Back