An Efficient Deep Learning Technique for Emotion Detection Using EEG Signal

EEG based emotion detection has been widely focused and it has applications in health care, HCI, and affective computing. In this paper, a new approach for detecting the emotion from the EEG signal is proposed by using Long Short-Term Memory (LSTM) networks. The proposed method is suitable for EEG data given temporal differences since it plays an important role when measuring emotional states. The first step is noise and artifacts reduction, the second step is extraction of features from time and frequency domain. All these constitute as a data input into an LSTM network, used for learning and outlining temporal characteristics inherent in the EEG sequences. and the performance of our approach is assessed on a standard EEG dataset with emotional annotations. The proposed LSTM-based model outperforms conventional ML methods and other deep learning models in terms of accuracy and their resistance to distortions.