EEG-Based Emotion Classification with Wavelet Entropy Feature

Traditional emotion recognition methods are mainly based on voice, expression and body movement. These physiological signals or facial expressions may hardly reveal inner emotions. In this paper, the wavelet entropy (WE) was utilized to represent the characteristics associated with emotional states. The average classification accuracies of positive, neutral and negative emotions are 70.65%, 70.53% and 70.28%, respectively. In order to demonstrate the effectiveness of the proposed method, the comparison experiments were carried out by the power spectrum density (PSD) feature and approximate entropy (ApEn) feature. The average classification accuracies are 70.49% (WE), 66.93% (PSD) and 64.44% (ApEn), respectively. The results indicate that the wavelet entropy feature performs better than the other two features for Electroencephalogram (EEG) based emotion recognition.