EEG-based emotion recognition has received increasing attention in advanced human-computer interaction, where the choice of independent variables to discriminate emotions from the frequency range of EEG and electrode locations is not very self-evident, thus this work tried to find the correlation between the emotional states and both EEG frequency ranges and EEG channels. 12 healthy volunteers were emotionally elicited by movie clips to experience five basic emotional states of neutral, happy, sad, tense and disgust states. Fisher discriminant ratio (FDR) was employed to find the discriminative bands and electrodes with statistical differences. Finally, a support vector machine (SVM) with 5-fold cross validation was performed. Average recognition rates have achieved 93.31% and 85.39% for two feature sets.
展开▼
