Emotion recognition based on multimodal physiological electrical signals

With the increasing severity of mental health problems, the application of emotion recognition techniques in mental health diagnosis and intervention has gradually received widespread attention. Accurate classification of emotional states is important for individual mental health management. This study proposes a multimodal emotion recognition method based on the fusion of electroencephalography (EEG) and electrocardiography (ECG) signals, aiming at the accurate classification of emotional states, especially for the three dimensions of emotions (potency, arousal, and sense of dominance). To this end, a composite neural network model (Att-1DCNN-GRU) is designed in this paper, which combines a one-dimensional convolutional neural network with an attention mechanism and gated recurrent units, and improves the emotion recognition by extracting the time-domain, frequency-domain, and nonlinear features of the EEG and ECG signals, and by employing a Random Forest approach to feature filtering, so as to improve the emotion recognition accuracy and robustness. The proposed model is validated on the DREAMER dataset, and the results show that the model achieves the three dimensions of emotion: value, arousal and dominance, with a high classification accuracy, especially on the 'value' dimension, with an accuracy of 95.95%. The fusion model significantly improves the recognition effect compared with the traditional emotion recognition methods using only EEG or ECG signals. In addition, to further validate the generalisation ability of the model, this study was also validated on the DEAP dataset, and the results showed that the model also performed well in terms of cross-dataset adaptation. Through a series of comparison and ablation experiments, this study demonstrates the advantages of multimodal signal fusion in emotion recognition and shows the great potential of deep learning methods in processing complex physiological signals. The experimental results show that the Att-1DCNN-GRU model exhibits strong capabilities in emotion recognition tasks, provides valuable technical support for emotion computing and mental health management, and has broad application prospects.