Recognition of Positive, Negative and Neutral Emotions Using Brain Connectivity Patterns

There are various resources inside the brain. Brain activities are the result of these sources or the result of their connectivity. Therefore, any special emotion should also be the result of various connectivity chains among the brain's resources. Studying this connectivity chains could help us recognize the corresponding emotions. The aim of this paper is to find interaction patterns in positive, neutral and negative emotions, and to recognize different types of emotions. We have used DEAP data in this project. These datasets were gathered from 32 volunteers, half of whom were women. Playing different types of music, caused them to experience special emotions, and their brain signals were recorded simultaneously. Music videos belonged to three different classes: positive, neutral and negative. After preprocessing the signals, we have achieved the connectional characteristics among the various channels, including causal features in various delays. Utilizing Davis-Bouldin Method, we obtained the sub-group of the optimal features. To evaluate the obtained results, we used SVM and KNN clustering methods. The final classified results, describes more favorable performance of interactional patterns and show the fact that connectional features can classify the classes in two arousal and valence with accuracy %79.7 and %88.2 respectively, which had %6 and %12.54 increase with respect to other traditional features.