Design of a radial basis function neural network for attention tasks event related potentials extraction

Abstract

Electroencephalogram (EEG) based biofeedback is widely employed to treat certain kinds of diseases especially Attention Deficit Hyperactivity Disorder (ADD/ADHD). Thus to design a system capable of learning a particular mapping between EEG features and different attention-level mental tasks is of great significance. Event Related Potentials (ERP) is such a powerful feature which is traditionally extracted by averaging. The paper proposed a new ERP extraction algorithm using radial basis function (RBF) neural network. It discussed the configuration, learning and running of the designed network. In order to reduce computational complexity and the influence of noise in estimating ERP, the partial least square regression was introduced to train the RBF network. Series experiments showed that the method is effective and is suitable for single-trail ERP estimation.