Stress Level Analysis Using EEG Signals: A Comparative Study of EEGNET and CNN Deep Learning Models

Understanding the brain's response to stress is pivotal for advancing mental health monitoring technologies. In this research paper, we leverage deep learning techniques to predict stress levels using electroencephalography (EEG) data from 10 students. This data captures brain activity during both rest and stress-induced conditions, with the stress induced through a visually demanding Stroop test. We focus on categorizing stress levels into four distinct categories: no stress, low stress, medium stress, and high stress. For analysis, we employ two advanced deep learning models, EEGNet and Convolutional Neural Network (CNN), which have demonstrated significant potential in managing the complexities of EEG data and identifying intricate patterns related to different stress levels.The methodology in this study includes rigorous data preprocessing, segmentation, and labeling to ensure the reliability and accuracy of the model training process. The EEG data from the 10 students are combined into a single dataset, with 70% allocated for training and 30% for testing. This approach ensures a robust evaluation of the models' predictive capabilities. We systematically compare EEGNet and CNN based on accuracy, precision, recall, and other relevant metrics. Our goal is to determine which model more effectively predicts stress levels and to understand their respective strengths and limitations. This comparative analysis provides valuable insights and paves the way for future applications in real-time stress monitoring.