Discrete Emotion Recognition using EEG Signals

Human machine interaction effectively improves the communication, which occurs between humans and machines. Electroencephalography (EEG) refers to a method which provides the monitoring of electrical activity in the brain with the help of electrical signals. Brain Computer Interface (BCI) refers to the system which converts the electrical signals which are produced by the brain to the signals which can be interpreted by a computer or an electronic system. The EEG signals are collected by the DEAP dataset, which refers to a Database for Emotion Analysis using Physiological signals. The second order butter worth filters preprocess the EEG data. The power of EEG can be taken by five frequency bands which are used as features. The P300 wave method is used for feature extraction. Mexican hat wavelet coefficients can be averaged across different scales for higher weighted features for classification. Out of 24 subjects of EEG data, 21 can be used for training and the rest can be used for testing. For classification of emotions, Support-Vector-Machine can be used as the tool. The system classifies the emotions in two different classes namely, Valence and Arousal dimensions. System classifies two un-pleasant emotions (anger, fear) and two-pleasant emotions (joy, amusement) and two with an average matching of 93.33%. But also classifies high-Arousal and low-Arousal with an average accuracy of 89.09%.