Abstract
This study presents an emotion detection system utilizing Electroencephalogram (EEG) data, integrated into the home robot's human-robot interaction model. With advancements in affective computing technology, the demand for home robots has been on the rise, offering emotional support and companionship through emotion recognition and responses, significantly enhancing the user experience. Leveraging the DEAP dataset, this study achieved real-time recognition of users' emotional states through preprocessing, feature extraction, and classification model training. We downsampled, removed artifacts from, and filtered the raw EEG signals, extracting features from the frequency, time, spatial domains, and brain networks. LASSO regression was employed for feature selection, followed by the application of various machine learning algorithms for emotion classification. The experimental results showed that the KNN classifier performed best in emotion recognition, achieving an average accuracy of 90.3%. Further human-robot interaction tests validated the system's practical applicability, with participants rating highly the precision of emotion detection, the timeliness of interactive responses, and the overall experience. The proposed EEG-based emotion recognition system shows significant potential in enhancing the human-robot interaction experience for home robots, while also suggesting directions for further improvements in real-time performance and personalized models.