Recurrent Neural Network Based Human Emotion Recognition using EEG brain signals

In daily lives, human experience a different emotion i.e.- sadness, angry, happiness. and rage that are very closely related tits behavior and play an important role in their daily routines. Emotion recognition means that identify and classify the human emotional mind state using their EEG brain signal, because EEG is very convenient and gives necessary information about brain emotion state. Since, emotion recognition through EEG signals is an active and challenging research area due to the non-stationary behavior of the signal. Therefore, Fastest, and accurate method is essential for automatically recognize and classify the human emotions. In this paper, we proposed RNN based Emotion Recognition model which consist of different layers i.e.- input layer, Dense layer, flatten layer, ReLU layer, SoftMax layer, and dropout layer that are used for feature extraction and model classification. ADAM optimization algorithm is used for model training which update the parameters of network after each iteration of epoch. EEG signals are used to train, validate and to predict the brain state of the human. The training accuracy of proposed model is 96.25% with 0.01 learning rate. To show the robustness and efficiency of proposed model, we compared with similar studies that shows performance of the proposed algorithm is better than the existing one in terms of better classification and model accuracy.