Ensemble Algorithms for EEG based Emotion Recognition

Abstract

Emotion recognition using Electroencephalogram (EEG) signal has grabbed the attention of researchers recently due to its widespread applications. This study employed empirical mode decomposition (EMD) to process EEG signals of different channel profiles and obtains various intrinsic mode functions. Sample Entropy (Samp En) is computed for the first four intrinsic mode functions, which are used as feature vectors for emotion recognition. To identify three categories of human emotions namely negative, neutral and positive, Random forest (RF) and Extreme Gradient Boosting (XGBoost) classifiers are fed with the extracted feature vectors. This algorithm achieved maximum accuracy of 88% and 96% with Random forest and XGBoost classifiers on a publicly available database SEED by considering all 62 channels of EEG.