Subject-independent Emotion recognition based on Entropy of EEG Signals

Emotion recognition is always an academic focus in the research of human-computer interaction (HCI). Due to poor generalizability of electroencephalogram (EEG) features from EEG different subjects, the aim of this paper is to further explore different characteristics in EEG signals to improve the accuracy of subject-independent emotion recognition. We extract the features from time domain and frequency domain, sample entropy and wavelet entropy and so on, and use SVM to evaluate the recognition performance. Results show the accuracy of public DEAP dataset is up to 70.1% in the valence dimension and 64.2% in the arousal dimension. This analysis is significant because we verify that the sample entropy and wavelet entropy are effective in subject-independent emotion recognition.