Causality Analysis of Emotional States from EEG Response

This paper proposes a novel study of the feature elements in corresponding emotional EEG states to investigate the appearance of repeating or residual elements which serve as the foundation of memory or causality in EEG signals. Features are first preprocessed and filtered to remove unwanted artifacts, and then are extracted by utilizing various well-known techniques including peak signal to noise ratio, Manhattan distance metric, mean cross correlation, mutual information and especially Granger causality. Once extracted, the host of features is used to perform statistical analysis on the EEG data. It is noteworthy that a key mechanism that has been used in this study is the consideration of only consecutive trials while performing statistical analysis for stronger establishment of our motive. Experimental results confirm that certain transition of emotional states such as fear to excitement and anger to sadness are more likely to be predictable given its past values than others.