Abstract: Emotions are produced by multiple channels in the brain, but when calculating the causal relationship between any two channels, the influence of other channels is often ignored. Therefore, the introduction of conditional Granger proposes a causal network emotion recognition method based on multi-channel EEG signals. The method uses conditional Granger to calculate the causal relationship of the whole brain channels under different emotions and construct a causal network based on the causal relationship, and analyze the in-degree and out-degree and betweenness topological properties of the channels in the causal network to find the key channel, and simplify the causality through the key channel. Finally, the causal connection relationship between nodes in the simplified causal network is used as a feature to train and classify under SVM and KNN. Experimental results show that the average recognition rates are 75.3% and 78.4%, respectively, and the proposed method is effective.