A Spatial-Temporal Feature Fusion Emotion Recognition Model Based on Attention Mechanism

Emotion recognition plays an import role in human-computer interaction, and electroencephalogram (EEG)-based emotion recognition has emerged as a reliable method extensively discussed by researchers. This paper introduces a novel spatial-temporal feature fusion model based on attention mechanism. The model consists of two distinct network branches. The first one utilizes the Temporal Convolutional Network (TCN) to capture temporal information from EEG data, while concurrently, the second branch employs the Convolutional Neural Network (CNN) to capture spatial information based on the functional connectivity matrix constructed from the raw EEG. Then, an attention-based feature fusion method is employed to combine the features produced by the two network branches. This paper presents experimental results based on the SEED-IV dataset, demonstrating the effectiveness of the attention-based feature fusion method, which significantly enhances the classification performance of the model.