A Brain Computer Interface Based on Neural Network with Efficient Pre-Processing

Brain computer interface (BCI) is one of hopeful interface technologies between human and machine. However, brain waves are very weak and there exist many kinds of noises. Therefore, what kinds of features are useful, how to extract the useful features, how to suppress noises, and so on are very important. On the other hand, neural networks are very useful technology for pattern classification. Especially, multilayer neural networks trained through the error back-propagation algorithm have been widely used in a wide variety of field. In this paper, the neural network is applied to the BCI. Amplitude of the FFT of the brain waves are used for the input data. Several kinds of techniques are introduced in this paper. Segmentation along the time axis for fast response, nonlinear normalization for emphasizing important information with small magnitude, averaging samples of the brain waves for suppressing noise effects and reduction in the number of the samples for achieving a small size network, and so on are newly introduced. Simulation was carried out by using the brain waves, which are available from the Web site of Colorado State University. The number of mental tasks is five. Ten data sets for each mental task are prepared. Among them, 9 data sets are used for training, and the rest one data set is used for testing. Selection of the one data set for testing is changed and accuracy of the correct classifications are averaged over the possible selections. Approximately, 80 % of correct classification of the brain waves is obtained, which is higher than the conventional