A New Feature Dimensionally Reduction Approach Based on General Tensor Discriminant Analysis in EEG Signal Classification

Feature selection from electroencephalogram (EEG) signals is important steps in BCI and medicine application. In this paper, a feature dimensionally reduction approach based on general tensor discriminant analysis (GTDA) is proposed. In this approach, EEG signal epochs are decomposed as spectral, spatial and temporal domain by Gabor functions as third order tensors. Then, projection vectors are extracted from tensor-represented EEG signals by GTDA. In this approach, the discriminative information in the training tensors is preserved that is a benefit in comparison with common feature space reduction approaches such as linear discriminant analysis (LDA) and principal component analysis (PCA). The proposed approach is evaluated to classify three mental tasks. The results indicate the improvement of classification performance in comparison with current methods.