EEG-Based Human Emotion Prediction Using an LSTM Model
Saeed Mohsen, Abdullah G. Alharbi
Abstract
It is deemed essential to identify and classify human emotions via deep learning with computers. Therefore, electroencephalogram (EEG) is extensively used as a physiological source of emotions. In this paper, a long short-term memory (LSTM) model is proposed for classification of positive, neutral, and negative emotions. This model is applied to a dataset that includes three classes of emotions with a total of 2,100 EEG samples from two subjects. The proposed model is trained using TensorFlow library with a tuning method to achieve maximum accuracy for emotion prediction. To appraise the model performance, receiver operating characteristic (ROC) curve is utilized. Experimental results demonstrate that the proposed model attains a high performance in the classification of human emotions. Furthermore, the proposed LSTM model has a testing accuracy of 98.13%, a macro average precision of 98.14%, and the area under the ROC curve for this model is 100%.