Research on Positive Emotion Recognition Based on EEG Signals

Human logical decision-making, perception, learning, and many functions, which influence people's decision-making on things and perception of external things, are all significantly influenced by emotions. An essential component of emotional computing is emotional cognition. By examining the psychological and psychological traits of the client, it evaluates the psychological condition of the service object. Currently, most of the research on emotions based on electroencephalogram (EEG) signals focuses on classifying positive, neutral, and negative emotions, or studying negative emotions, with less attention paid to specifically identifying positive emotions. This experiment proposed an experimental design that utilized virtual reality technology as an inducing method to stimulate positive emotions, and identified and evaluated the emotions of happiness, desire, and healing. This experiment collected datasets on three positive emotions of happiness, desire, and healing, and collected the Positive Affect and Negative Affect Scale (PANAS) and self-assessment (SAM) form of the participants. Divided the experiment into two groups, one was the experimental group wearing VR, and the other was the experimental group not wearing VR. The immersion feeling scale was used to study the effect of VR on stimulating emotions. Through the design of emotion induction experiments, the EEG signals of experiencing happiness, desire, and healing were collected. The EEG signals were input into a CNN network for feature extraction, both in the form of images and time series. The Resnet18 network was used for image-based emotion recognition with an accuracy of 93%. The time-series data was processed using an LSTM network for emotion recognition with an accuracy of 94.9%.