A Long Short Term Memory Deep Learning Network for the Classification of Negative Emotions Using EEG Signals

Abstract

In cognitive science and human-computer interac-tion, automatic human emotion recognition using physiologicalstimuli is a key technology. This research considers classificationof negative emotions using EEG signals in response to emotionalclips. This paper introduces a long short term memory deeplearning (LSTM) network to recognize emotions using EEGsignals. The primary goal of this approach is to assess theclassification performance of the LSTM model for classifyingemotions. The secondary goal is to assess the human behavior ofdifferent age groups and genders. We have compared the per-formance of Multilayer Perceptron (MLP), K-nearest neighbors(KNN), Support Vector Machine (SVM), Deep Belief Networkbased SVM (DBN-SVM), and LSTM based deep learning modelfor classification of negative emotions using brain signals. Theanalysis shows that for four class of negative emotion recognitionLSTM based deep learning model provides classification accuracyas 81.63%, 84.64%, 89.73%, and 92.84% for 50-50, 60-40, 70-30,and 10-fold cross-validation. Generalizability and reliability ofthis approach is evaluated by applying our approach to publiclyavailable EEG dataset DEAP and SEED. In compliance with theself-reported feelings, brain signals of 26-35 years of age groupprovided the highest emotional identification. Among genders,females are more emotionally active as compared to males.