Human emotion is a physical and physiological activity that is triggered in response to internal or external stimulation. Automating emotion recognition is a challenging task since emotions change continuously with time, situation, etc. It is a complex activity. Recently, multiple studies were undertaken using CAD tools to automatically detect emotions. EEG signals are one mode through which emotions can be captured. These signals are captured using multiple channels. Processing all these channel features can be a cumbersome task. Hence in the current study, we tried to recognize the emotions using a single T7 channel. Initially, the features of the pre-processed EEG channel data are extracted using continuous wavelet transform (WT). Further, the dimensionality of the matrix is reduced by principal component analysis. The grey wolf optimization (GWO) is used for optimization purposes. The trained features are classified using linear SVM, quadratic SVM, and a wide neural network. A wide neural network attained the maximum accuracy of 88.10%. Further, the study can be extended to study the significance of other channel features in emotion recognition.
Gudigar, Anjan; U, Raghavendra Dr.; M, Maithri; and Praharaj, Samir Kumar Dr
"Wavelet-based Single-Channel EEG Features for the Automated Recognition of Human Emotion,"
Manipal Journal of Science and Technology: Vol. 6:
1, Article 6.
Available at: https://impressions.manipal.edu/mjst/vol6/iss1/6