As a long-standing research topic in the field of brain–computer interface, emotion recognition still suffers from low recognition accuracy. In this research, we present a novel model named DE-CNN-BiLSTM deeply integrating the complexity of EEG signals, the spatial structure of brain and temporal contexts of emotion formation. Firstly, we extract the complexity properties of the EEG signal by calculating Differential Entropy in different time slices of different frequency bands to obtain 4D feature tensors according to brain location. Subsequently, the 4D tensors are input into the Convolutional Neural Network to learn brain structure and output time sequences; after that Bidirectional Long-Short Term Memory is used to learn past and future information of the time sequences. Compared with the existing emotion recognition models, the new model can decode the EEG signal deeply and extract key emotional features to improve accuracy. The simulation results show the algorithm achieves an average accuracy of 94% for DEAP dataset and 94.82% for SEED dataset, confirming its high accuracy and strong robustness.

EEG-based emotion recognition is becoming crucial in brain-computer interfaces (BCI). Currently, most researches focus on improving accuracy, while neglecting further research on the interpretability of models, we are committed to analyzing the impact of different brain regions and signal frequency bands on emotion generation based on graph structure. Therefore, this paper proposes a method named Dual Attention Mechanism Graph Convolutional Neural Network (DAMGCN). Specifically, we utilize graph convolutional neural networks to model the brain network as a graph to extract representative spatial features. Furthermore, we employ the self-attention mechanism of the Transformer model which allocates more electrode channel weights and signal frequency band weights to important brain regions and frequency bands. The visualization of attention mechanism clearly demonstrates the weight allocation learned by DAMGCN. During the performance evaluation of our model on the DEAP, SEED, and SEED-IV datasets, we achieved the best results on the SEED dataset, showing subject-dependent experiments’ accuracy of 99.42% and subject-independent experiments’ accuracy of 73.21%. The results are demonstrably superior to the accuracies of most existing models in the realm of EEG-based emotion recognition.

To describe both the global and local characteristics of a network more comprehensively, we propose the weighted K-order propagation number (WKPN) algorithm to extract the disease propagation based on the network topology to evaluate the node importance. Each node is set as the source of infection, and the total number of infected nodes is defined as the K-order propagation number after experiencing the propagation time K. The simulation of the symmetric network with bridge nodes indicated that the WKPN algorithm was more effective for evaluation of the algorithm features. A deliberate attack strategy, which indicated an attack on the network according to the node importance from high to low, was employed to evaluate the WKPN algorithm in real networks. Compared with the other methods tested, the results demonstrate the applicability and advancement that a lower number of nodes, with a higher importance calculated by the K-order propagation number algorithm, has to achieve full damage to the network structure.

Working Memory (WM) is a short-term memory for processing and storing information. When investigating WM mechanisms using Electroencephalogram (EEG), its rhythmic synchronization properties inevitably become one of the focal features. To further leverage these features for better improve WM task performance, this paper uses a novel algorithm: Weight K-order propagation number (WKPN) to locate important brain nodes and their coupling characteristic in different frequency bands while subjects are proceeding French word retaining tasks, which is an intriguing but original experiment paradigm. Based on this approach, we investigated the node importance of PLV brain networks under different memory loads and found that the connectivity between frontal and parieto-occipital lobes in theta and beta frequency bands enhanced with increasing memory load. We used the node importance of the brain network as a feature vector of the SVM to classify different memory load states, and the highest classification accuracy of 95% is obtained in the beta band. Compared to the Weight degree centrality (WDC) and Weight Page Rank (WPR) algorithm, the SVM with the node importance of the brain network as the feature vector calculated by the WKPN algorithm has higher classification accuracy and shorter running time. It is concluded that the algorithm can effectively spot active central hubs so that researchers can later put more energy to study these areas where active hubs lie in such as placing Transcranial alternating current stimulation (tACS).