Research on Classification and Recognition of Six Basic Micro-expressions Based on EEG Signals

Currently, Micro-expression recognition (MER) holds significant application value. However, due to the brief duration and low intensity of MEs, accurately and quickly recognizing them remains a challenging problem. Electroencephalogram (EEG) signals, with their high temporal resolution, can capture immediate responses at the moment MEs occur. Therefore, we propose a novel method based on EEG for the fast and accurate recognition of MEs. First, we constructed the EEG-based ME dataset SWU-ME by combining OpenFace 2.0 software and the Facial Action Coding System (FACS). Next, we developed a neural network model (ME-SECN) based on frequency band weighting and spatial information extraction from EEG channels. Experimental validation shows that, compared to existing advanced EEG recognition methods, the ME-SECN model performs best in recognition accuracy, achieving an accuracy of 90.70% and an F1 score of 90.78%. The results demonstrate that EEG signals can effectively classify and recognize MEs, and the proposed model can more efficiently identify different types of MEs.