An Optimal EEG-based Emotion Recognition Algorithm Using Gabor Features

Feature extraction and accurate classification of the emotion-related EEG-characteristics have a key role in success of emotion recognition systems. In this paper, an optimal EEG-based emotion recognition algorithm based on spectral features and neural network classifiers is proposed. In this algorithm, spectral, spatial and temporal features are selected from the emotion-related EEG signals by applying Gabor functions and wavelet transform. Then neural network classifiers such as improved particle swarm optimization (IPSO) and probabilistic neural network (PNN) are developed to determine an optimal nonlinear decision boundary between the extracted features from the six basic emotions (happiness, surprise, anger, fear, disgust and sadness). The best result is obtained when Gabor-based features and PNN classifier are used. In this condition, our algorithm can achieve average accuracy of 64.78% that can be used in brain-computer interfaces systems. Key-Words: electroencephalogram, emotion recognition, wavelet transform, Gabor functions, improved particle swarm optimization (IPSO), probabilistic neural network (PNN)