EEG-based classification of emotions using empirical mode decomposition and autoregressive model

Emotion can be classified based on 2-dimensional valence-arousal model which includes four categories of emotional states, such as high arousal high valence, low arousal high valence, high arousal low valence, and low arousal low valence. In this paper, we present the attempt to investigate feature extraction of electroencephalogram (EEG) based emotional data by focusing on empirical mode decomposition (EMD) and autoregressive (AR) model, and construct an EEG-based emotion recognition method to classify these emotional states. We first employ EMD method to decompose EEG signals into several intrinsic mode functions (IMFs), and then the features are calculated from IMFs based on AR model using a sliding window, and finally we use these features to recognize emotions. The average recognition rate of our proposed method is 86.28% for 4 binary-class tasks on DEAP dataset. Experimental results show that our proposed method has a uniform and stable performance of emotion recognition, which are quite competitive with the results of methods of comparison.