Power Spectral Density Features for Classifying Action Intention Understanding EEG Signals

Abstract

Background: Classification of action intention understanding is extremely important for social interaction and brain-computer interface (BCI). However, it is very difficult to obtain a satisfactory experimental result. Method: This study first extracts power spectral density (PSD) features based on preprocessed EEG signals, and then selects the effective features by statistical thresholds. Results: Under different combining conditions from three pairwise action intention stimuli and five frequency bands, some electrodes show manifest statistical differences, as well as most of subjects obtain high average classification accuracies. Conclusions: The PSD features selected with statistical thresholds are exceedingly useful for the classification task of action intention understanding EEG signals.