Electroencephalogram Emotion Recognition Based on A Stacking Classification Model

To improve the accuracy of Electroencephalogram (EEG) emotion recognition, a stacking emotion classification model is proposed, in which different classification models such as XGBoost, LightGBM and Random Forest are integrated to learn the features. In addition, the Renyi entropy of 32 channels' EEG signals are extracted as the feature and Linear discriminant analysis (LDA) is employed to reduce the dimension of the feature set. The proposal is tested on the DEAP dataset, and the EEG emotional states are accessed in Arousal-Valence emotion space, in which HA/LA and HV/LV are classified, respectively. The result shows that the average recognition accuracies of 77.19% for HA/LA and 79.06% for HV/LV are obtained, which demonstrates that the proposal is feasible in EEG emotion recognition.