چكيده به لاتين
Abstract:
In the current era, implicit tagging of multimedia contents is going to reach a new level through availability of physiological information (e. g. electroencephalogram (EEG) signals). Thus implicit tagging is main idea in emotion recognition. Brain as the area calls emotion activity contain the best affective information. The physiological signals shown that the EEG has a lot of information in relation to the emotional states of users. Emotion is a psychophysiological process that plays an important role in human communication and current efforts in Human-Machine-Interaction (HMI). This project aims at finding the relationship between functional connectivity of EEG signals and human emotions. In the realm of affective computing, recognition of emotional state of users has significantly been essential. This study applied pattern recognition algorithms to separate emotional state of the subjects. To this end, we propose an emotion recognition system based on EEG connectivity features between electrodes to classify two dominant emotion model based on valence and arousal, during video stimuli presentation. EEG features were extracted using the Pearson’s correlation coefficient (Corr), Phase Locking Value (PLV), Phase-Lag Index (PLI), Weighted Phase-Lag Index (wPLI),Mutual Information (MI) and Magnitude square coherence estimation of the pair electrodes of EEG signals. We selected the robust features through Fisher linear discriminant analysis method. Leave one out cross-validation was then performed using SVM classifier to classify the two emotional states (high/low Arousal, high/low Valence).A open access affective dataset for emotion analysis using EEG, physiological and video signals (DEAP) is used for evaluation of the proposed system. In this database 40 audio-visual stimuli have been used. The EEG data collected from 32 subjects during stimulus presentation. The subjects self-reported the level of their arousal, valence, dominance, and liking after each clip. The results showed that the proposed method improved the average classification accuracies for valence and arousal in compared to traditional features used for this aim. The results indicate that it is possible to use functional connectivity features for noninvasive assessment of the emotional states with high accuracy. Therefore ,functional connectivity features as useful features can be used to detect emotion.
Keywords: Emotion, EEG, Functional connectivity, Affective classification
