EEGNetT: EEG-based neural network for emotion recognition in real-world applications

In this study, we propose an electroencephalogram (EEG)-based emotion recognition (ER) system with a newly developed network, EEGNetT. In addition to the recognition accuracy, some indicators, including the scale of the training dataset, the number of network parameters, and the time consumption of convergence were taken into account for the evaluation of the system performance. The results obtained for the ER task indicated that EEGNetT achieves superior performance than EEGNet, especially with small training datasets. Additionally, the number of network parameters and the average epoch time indicate that EEGNetT is a promising network for real-world applications.