A real-time model based Support Vector Machine for emotion recognition through EEG

Recently, there has been a significant amount of work on the recognition of human emotions. The results of the work can be applied in real applications, for example in market survey or neuro-marketing. This interesting problem requires to recognize naturally human emotions which come from our mind but ignore the external expressions fully controlled by a subject. A popular approach uses key information from electroencephalography (EEG) signals to identify human emotions. In this paper, we proposed an emotion recognition model based on the Russell's circumplex model, Higuchi Fractal Dimension (HFD) algorithm and Support Vector Machine (SVM) as a classifier. Moreover, we also proposed a method to determine an emotion label of a series of EEG signals. Our model includes two main approaches in machine learning step. In a first approach, machine learning was utilized for all EEG signals from numerous subjects while another used machine learning for each particular subject. We extensively implemented our model in several test data. The experimental results showed that the first approach is impossible to apply in practical applications because EEG signal of each subject has individual characteristic. In addition, in the second, our model can recognize five basic states of human emotion in real-time with average accuracy 70.5%.