Emotion Recognition Using the Fusion of Frontal 2-channel EEG Signals and Peripheral Physiological Signals

Emotion states are a direct response of human beings to the state of the external environment. Emotion recognition based on EEG signals is currently one of the important technical methods in the field of affective computing. However, the excessive number of EEG channels may lead to high dimensions and large computing time. To solve this problem, we study the feasibility of using EEG channels in the hairless area of the prefrontal to recognize emotions. Furthermore, we innovatively propose to use peripheral physiological signals to supplement the missing EEG channels for emotion recognition. Then we use the DEAP dataset to verify the feasibility of the proposed method. The results show that the 2-channel EEG signals acquired from FP1 and FP2 position can achieve comparable accuracy and F1 score than that of using 18-channel EEG signals. Furthermore, the feature matrix fused the peripheral physiological has achieved better accuracy and F1 score than that of using 18-channel EEG signals. In addition, the training time can be greatly decreased by the fused feature matrix. The research in this paper is of great significance to the future development of emotion recognition based on multi-modal physiological signals acquired by wearable devices.