Dream Emotion Prediction Using Artificial Intelligence: A Comparative Analysis with Eeg Signals

Abstract

This research paper seeks to deliver an in-depth exploration of the practical uses of artificial intelligence in forecasting emotional states during dreams through the analysis of EEG signals. It specifically evaluates the effectiveness of four machine learning models—Random Forest (RF), Support Vector Machines (SVM), Neural Networks (NN), and Gradient Boosting (GB)—in categorizing emotions derived from EEG data collected during REM sleep, a phase most closely linked to vivid dreaming (1). The dataset comprises EEG signals from various participants, with emotional states classified as Happy, Sad, Neutral, Angry, Calm, and Fearful. The models are assessed based on metrics such as accuracy, precision, recall, F1-score, and AUC-ROC. Findings indicate the promise of AI in interpreting EEG signals to anticipate emotions during dreams, with Random Forest attaining the highest accuracy and F1-score. To strengthen the validity of the results, cross-validation was conducted, and the models were optimized using grid search methods. Techniques for feature extraction, including power spectral density and wavelet transform, were utilized to enhance model efficacy by isolating frequency-domain features associated with emotional states. This study underscores the importance of utilizing EEG biomarkers for emotion recognition, offering insights into subconscious emotional processing (2). The application of AI in this field could pave the way for innovative diagnostic tools in mental health, such as early identification of emotional disorders, tailored therapy, and a deeper understanding of the emotional processing related to dreams.