Emotion Recognition Using Time-frequency Analysis of EEG Signals and Machine Learning

In recent years, emotion recognition has drawn intense interest from researchers in various fields. Because of their intrinsic correlation with emotions, physiological signals based emotion recognition method is objective and insusceptible to intentional disguise of the subject under study. In particular, electroencephalogram (EEC) signals arc known to be responsive and sensitive to variations in emotional state. In this paper, a 4-class emotion classification problem is investigated. Firstly, a clustering algorithm is used to determine the target class of each emotion-related data. Then wavelet analysis is used to extract features from 32-channel EEC signals. Finally, we compare live feature dimensionality reduction or feature selection algorithms and four types of machine learning based classifiers. The comprehensive comparative results show the effectiveness of the combination of EEC feature dimensionality reduction algorithm and random forest (RF) for multi-class emotion recognition problem under study.