Classification of EEG signals for different emotional states

We propose a recognition system by using electroencephalogram (EEG) data for emotion classification. Keeping in mind the growing interest for emotion detection automatically, we are trying to identify the prominent brain waves (i.e. alpha, beta, delta, theta) for particular emotion. For analysis we have used frequency data. EEG data was collected by showing and playing different audio-video stimuli to acquire the proper emotions. Detailed analysis of the dominating signals when exposed to different emotions (happy/sad) was conducted. We made use of standard statistical techniques for feature extraction. It is found that when people are exposed to specific emotion like happiness or sadness, higher frequency signals are more prominently seen as compared to lower frequency signals, in particular regions of the brain. During intense emotional activity, changes were noticed in the alpha signal in occipital and frontal regions of the brain. In case of very intense sad emotion display, Beta signals were also seen over Temporal and Frontal regions. For classification of data we have used Linear Discriminant Analysis (LDA). The classification rate in case of sad emotions is 84.37%, for happiness it is 78.12% and for relaxed state it is found to be 92.70%.