Stress Classification Using Physiological Signals: A Comprehensive Review of Methods and Approaches Combined With a Novel CNN-Based Ecg Experiment

Accurate stress detection through physiological signals holds significant promise for improving healthcare outcomes, reducing costs, and enabling early intervention in stress-related disorders. This study presents a comprehensive review of recent advances in stress classification using physiological data, highlighting key methods, challenges, and emerging trends in the field. Particular emphasis is placed on the limitations posed by small datasets, the importance of personalized models, and the difficulties of real-time application in uncontrolled environments. In parallel, we propose and evaluate a novel convolutional neural network (CNN) architecture designed to classify electrocardiogram (ECG) signals into four distinct categories. The model demonstrates robust learning and moderate generalization under data-constrained conditions, achieving 60.95% accuracy on an independent test set. The findings reinforce the efficacy of deep learning in stress classification and underscore the necessity for personalized, real-time, and multimodal approaches in future research.