An Empirical Study on EEG Signals for Emotion Recognition Using SEED

Abstract

Human emotions are too complex to be accurately recognized by others. In the era of Artificial Intelligence (AI), automatic emotion recognition has become an active field for research and applications.  The technology of both AI could have a significant impact on public health. There are a variety of scientific methods that can precisely measure emotions even in the face of impassivity. Some of the most reliable methods include the electroencephalography (EEG) which depends on physiological signals. EEG-based emotion recognition has received much exploration in recent years. The SJTU Emotion EEG Dataset SEED is an open-source dataset that contains EEG signals used for emotion recognition. Most EEG-based emotion recognition research applies machine learning techniques for classifying emotions. In this paper, we conduct an empirical study on SEED dataset to investigate some characteristics of this dataset. We find that the recorded emotions among multiple sessions are the same for most participants. In addition, there is a difference in the detected emotions between participants from the same gender. Finally, the emotions between biologically male and female participants are distinctive