Search

Emotion recognition is one of the most researched fields in today's scientific world. It is a subject that is closely examined by disciplines such as neuroscience and psychology, and it is more and more involved in our daily lives, especially in human-computer interaction area. Although methods such as speech signals, facial expressions, body language, and facial expressions are used for emotion analysis, these methods do not give as reliable results as biological signals because they are open to manipulation. In this study, a new method for emotion recognition with electroencephalography (EEG) signals, which is a bioelectrical signal prepared with the help of virtual reality (VR) technology, is proposed. In this method, differential entropy (DE) and phase-locking value (PLV) properties of sub-bands of EEG signals were used to recognize positive and negative emotions with the help of a designed one-dimensional convolutional neural network (1D-CNN). The feature matrices obtained with the help of both features were tested ten times and average accuracy values were obtained. As a result of these tests, the highest average accuracy scores with DE and FKD features were obtained as 74.061±1.41% and 63.7590±1.72%, respectively, by combining all sub-band feature matrices. In addition, the higher accuracy rates of the tests of the high-frequency signal components obtained in the study compared to the low-frequency bands, supported the results of similar studies carried out in this area before. [ABSTRACT FROM AUTHOR]