Emotion recognition based on electroencephalograms has become an active research area. Yet, identifying emotions using only brainwaves is still very challenging, especially the subject-independent task. Numerous studies have tried to propose methods to recognize emotions, including machine learning techniques like convolutional neural network (CNN). Since CNN has shown its potential in generalization to unseen subjects, manipulating CNN hyperparameters like the window size and electrode order might be beneficial. To our knowledge, this is the first work that extensively observed the parameter selection effect on the CNN. The temporal information in distinct window sizes was found to significantly affect the recognition performance, and CNN was found to be more responsive to changing window sizes than the support vector machine. Classifying the arousal achieved the best performance with a window size of ten seconds, obtaining 56.85% accuracy and a Matthews correlation coefficient (MCC) of 0.1369. Valence recognition had the best performance with a window length of eight seconds at 73.34% accuracy and an MCC value of 0.4669. Spatial information from varying the electrode orders had a small effect on the classification. Overall, valence results had a much more superior performance than arousal results, which were, perhaps, influenced by features related to brain activity asymmetry between the left and right hemispheres.

Emotion recognition during music listening using electroencephalogram (EEG) has gained more attention from researchers, recently. Many studies focused on accuracy on one subject while subject-independent performance evaluation was still unclear. In this paper, the objective is to create an emotion recognition model that can be applied to multiple subjects. By adopting convolutional neural networks (CNNs), advantage could be gained from utilizing information from electrodes and time steps. Using CNNs also does not need feature extraction which might leave out other related but unobserved features. CNNs with three to seven convolutional layers were deployed in this research. We measured their performance with a binary classification task for compositions of emotions including arousal and valence. The results showed that our method captured EEG signal patterns from numerous subjects by 10-fold cross validation with 81.54% and 86.87% accuracy from arousal and valence respectively. The method also showed a higher capability of generalization to unseen subjects than the previous method as can be observed from the results of leave-one-subject-out validation.