EEG-based subject-dependent emotion recognition algorithm using fractal dimension

In this paper, a real-time Electroencephalogram (EEG)-based emotion recognition algorithm using Higuchi Fractal Dimension (FD) Spectrum is proposed. As EEG is a nonlinear and multi-fractal signal, its FD spectrum can give a better understanding of the nonlinear property of EEG. Three values are selected from the whole spectrum and are combined with the other features such as statistical and Higher Order Crossings ones. The Support Vector Machine is used as the classifier. The proposed algorithm is validated on both benchmark database DEAP with video stimuli and our own dataset which used visual stimuli to evoke emotions. Up to 8 emotions can be recognized with only 4 channels. The experiment analysis results show that using FD spectrum features it is possible to improve classification accuracy.