SELF-SUPERVISED LEARNING FOR ACUTE STRESS DETECTION IN ELECTROCARDIOGRAMS

This project utilizes advancements in ECG technology and deep learning to improve stress detection using raw ECG data. There is an anticipated significant increase in the use of electrocardiograms in the near future. Advancements in recording technology have enabled non-medical applications, allowing for fewer, or even no, electrodes with radar- based approaches. Concurrently, the computational revolution, led by transformers, has fundamentally changed the processing of sequential data. In addition, self-supervised learning has emerged as a powerful strategy to utilize vast amounts of unlabelled ECG data, enhancing the robustness and performance of our stress detection models. This synergy sets the stage for more reliable, stress detection models, enhancing our understanding of the impact of stress on driving and potentially informing safety measures.