A Decision Science Approach Using Hybrid EEG Feature Extraction and GAN-Based Emotion Classification

[Purpose] Emotions play an essential role in human life and they profoundly influence behavior, decision-making, and well-being. This approach aims to classify human emotions by using Generative Adversarial Networks (GAN) and hybrid Electroencephalography (EEG) features with the DEAP dataset. [Design/methodology/approach] The proposed system addresses the limitations of traditional classification techniques by generating synthetic hybrid features that capture additional information about emotional states. Informed by decision science principles, the system recognizes that emotions heavily influence human decision-making processes. [Findings] The process consists of data collection, pre-processing, feature extraction, GAN training, hybrid feature generation, and classification. The DEAP dataset is pre-processed by using Independent Component Analysis (ICA) and Wavelet Transform to remove artifacts. A GAN model is trained to generate synthetic features that mimic the distribution of real EEG signals. The hybrid features are generated by combining the real EEG features and synthetic features. [Originality/value] The performance of the classification system is evaluated using accuracy at 97.4%, precision at 97.22%, recall at 96.8%, and F1 score at 97.08%. By leveraging EEG signals, the proposed system shows promise in enhancing the accuracy of emotion classification, opening up exciting avenues for future research in this domain.