Cascaded Convolutional Recurrent Neural Networks for EEG Emotion Recognition Based on Temporal-Frequency-Spatial Features

Emotion recognition is a research area that spans multiple disciplines, including computational science, neuroscience, and cognitive psychology. The use of electroencephalogram (EEG) signals in emotion recognition is particularly promising due to their objective and nonartefactual nature. To effectively leverage the spatial information between electrodes, the temporal correlation of EEG sequences, and the various sub-bands of information corresponding to different emotions, we construct a 4D matrix comprising temporal-frequency-spatial features as the input to our proposed hybrid model. This model incorporates a residual network based on depthwise convolution (DC) and pointwise convolution (PC), which not only extracts the spatial-frequency information in the input signal, but also reduces the training parameters. To further improve performance, we apply frequency channel attention networks (FcaNet) to distribute weights to different channel features. Finally, we use a bidirectional long short-term memory network (Bi-LSTM) to learn the temporal information in the sequence in both directions. To highlight the temporal importance of the frame window in the sample, we choose the weighted sum of the hidden layer states at all frame moments as the input to softmax. Our experimental results demonstrate that the proposed method achieves excellent recognition performance. We experimentally validated all proposed methods on the DEAP dataset, which has authoritative status in the EEG emotion recognition domain. The average accuracy achieved was 97.84% for the four binary classifications of valence, arousal, dominance, and liking and 88.46% for the four classifications of high and low valence-arousal recognition.