Exploring Humanoid Robot Face Preference Using Brain Functional Connectivity and Graph Neural Network

Graph neural networks (GNN) have been applied in EEG signal analysis. However, it is not clear how to describe the connection relationships between electrodes, and a reasonable representation of the adjacency matrix cannot be neglected for the study of GNN based on EEG signals. In this paper, we use brain functional connectivity to analyze the face preferences of a humanoid robot and explore the feasibility of using the matrix obtained based on brain functional connectivity measurements as a GNN connectivity matrix. The results show that it is feasible and an average accuracy of 73.47 ^% is achieved. In addition, the results of the brain functional connectivity analysis also show that there is a difference between preferences and non-preferences.