Classification of EEG Signals Variation in Emotional State Using Higher Order Spectral

Emotions play an important role in human life. Electroencephalogram (EEG) is the reflection of brain activity and is widely used in clinical diagnoses and biomedical researches. This paper proposes an emotion recognition system based on EEG signals. The visual-induction-based acquisition protocol has been designed for acquiring the EEG signals under three emotional states (calm-neutral, positive-excited and negative-excited) for participants. After preprocessing the signals, higher order spectra (HOS) are employed to extract the many features required for classifying human emotions. In this paper, for the first time, two values are used, such as bispectrum and bicoherence, in order to evaluate the EEG changes in the excited state. In this work, we used genetic algorithm and support vector machines (SVM) for classifying the emotions. The results show that, most EEG signal variations correlated with beta frequency band. Hence, with selected HOS-based features, we achieve average accuracy of 52% for three Categories. The results confirm the possibility of using HOS-based features in assessing human emotions from EEG signals. In comparison to other works, they did not show any significant difference in accuracy.HOSis auseful method in representation of emotional states of thebrain.