EEG Emotion Recognition Based on 3D-CTransNet.
Journal:
Annual International Conference of the IEEE Engineering in Medicine and Biology Society. IEEE Engineering in Medicine and Biology Society. Annual International Conference
PMID:
40031451
Abstract
Emotion recognition is of great significance for brain-computer interface and emotion computing, and EEG plays a key role in this field. However, the current design of brain computer interface deep learning model is faced with algorithmic or structural constraints, and it is difficult to recognize the complex features in EEG signals with long-term dynamic changes. To solve this issue, a hybrid CNN-Transformer structure using 3D data input is proposed and named 3D-CTransNet in this paper, which solves the problem of performance degradation of the traditional CNN-LSTM hybrid structure in the recognition of long sequence signals. At the same time, the self attention mechanism and parallel mode introduced by Transformer improve the recognition accuracy and processing speed. In addition, the 3D data feature map based on electrode position mapping effectively retains the spatial characteristics of EEG signals, which makes CNN better combine the time domain and spatial domain. Finally, the Valence-Arousal classification training of emotion is carried out on the public dataset DEAP, and the classification accuracy is 97.04%, which is about 5% higher than that of the hybrid CNN-LSTM model.