EEG-based emotion recognition utilizing wavelet coefficients

This paper focuses on EEG (Electroencephalography) signals as a robust method for emotion recognition. In emotion recognition, researchers usually use features such as eye pupil diameter, facial features, EEG signals and physiological signals like: respiration amplitude, heart rate, skin temperature, blood volume pulse, respiration rate etc. In this paper we use just EEG signals as we believe that a human being may suffer from some physical disabilities and impairments like visual disorders, motor impairment or some other common disorders. So, the use of EEG signal, in some aspects, can be more useful and utilizable in real life. In this paper, we use a combination of some existent techniques on this theme, such as wavelet coefficients and an 8-number electrode configuration, which makes our approach really convenient and comfortable to use, and some other methods that may seem minor; But the way we employ and combine them, make a novel, productive, high efficient and reliable algorithm that highly can help people with some special disorders. To have a brief overview of the results of our work: the average Arousal F-Score and Valence F-Score for our algorithm are, respectively, 0.73 and 0.77. These values for a corresponding work are, 0.60 and 0.50, respectively. As it is seen the results have improved by 0.13 and 0.27. The results of our EEG-based algorithm are even better than the fusion of facial and EEG signals or physiological signals presented in the corresponding works. Beside this better performance, the ease and comfort that our method provides for users, is far beyond description.