Spatio-temporal CNN-BiLSTM dynamic approach to emotion recognition based on EEG signal

In this paper, a hybrid CNN-BiLSTM model for EEG-based emotion detection system is presented. The proposed technique is developed by extracting features using Power Spectral Density (PSD) signal. The proposed approach is carried out by combining CNN and bidirectional LSTM models to increase the comprehension of context in sequential data. The proposed approach is tested on the widely-used SEED datasets for the accurate classification of milder emotions such as positive, negative and neutral. The proposed approach is designed with the effectiveness in extracting spatial features of CNN architecture and LSTM network are utilized for their capability in modeling temporal relationships in EEG signals. The proposed approach is robust because experimentally the proposed approach yields a rate of 97.5 % accuracy to categorize emotions, improving the performance of EEG-based emotion recognition systems, opening up new possibilities for developing advanced brain monitoring and real-time emotion-aware systems.