基于脑电图的增强型图卷积网络架构的情感识别OA

In view of the affective computing problem of existing machine learning models in different electroen-cephalogram channels,different frequency bands within the same channel,and individual differences among different subjects,this study proposes an enhanced GCN(Graph Convolutional Network)based on EEG(Electroencephalo-gram)for affective recognition.The potential complementary information is explored through the multi-head attention mechanism,and the problem of insufficient processing of the importance differences of different electroencephalogram channels and frequency bands in existing emotion recognition models is solved by the fusion of the adjacency matrix and the extension matrix.The problems of differences in brain structure and activity patterns in cross-subject tasks are improved through adversarial training methods.The experimental results show that under the two paradigms of test-de-pendent and test-independent,the classification accuracies of the proposed method in the SEED-IV dataset are 84.46%and 75.92%respectively,and in the SEED-V dataset are 81.08%and 66.05%respectively.