Adaptive neural network classifier for EEG signals of six mental tasks

In this paper, a new adaptive neural network classifier of six different mental tasks from EEG-based P300 signals is proposed. To overcome the classifier of overtraining caused by noisy and non-stationary data, the EEG signals are filtered and extracted using autoregressive model before passed to the adaptive neural network classifier. To test the improvement in the EEG classification performance with the proposed method, comparative experiments were conducted using Bayesian Linear Discriminant Analysis. All subjects achieved a classification accuracy of 100%.