Exploring the impact of Stress and Cognitive Workload on Eye Movements: A Preliminary Study

The evolution of technology used in the service of human work has led to the development of human-machine interaction (HMI) systems. They enable the performance of certain functions and the achievement of specific goals through collaboration between humans and machines. These systems are incredibly innovative and can be extremely useful for humans; moreover, the HMI systems can be used mainly in two ways. In some cases, they are specifically designed to relieve the subject of several tasks carried out directly by the machine. In other cases, the operator must fulfill several duties simultaneously because of the interaction with the HMI device. These situations generate high stress and mental workload (MWL) levels in the subject, leading him or her to perform incorrect actions and risky situations in which accidents are more likely to occur. Hence, HMI systems must be created to avoid these dangerous situations for the person interacting with the machine and those around him. These safety mechanisms are also based on measuring and controlling the subjects’ stress and cognitive load, preventing them from performing tasks or functions if these two altered states are too pronounced. In the context of this study, reference is made to aviation pilots who increasingly interact with machines during their work. There is also a tendency to move from having two pilots per aircraft to only one by replacing the copilot with an HMI system; the so-called Single Pilot Operation SPO. HMI technologies have the task of supervising the pilots and, in case these systems notice that they are overloaded, giving support by simplifying the charges they must fulfill or even replacing the pilots themself. Various biosignals have been seen to visualize changes in these two altered states; the most studied are the f-NIRS, the EDA, the ECG, the respiration, the body temperature, and the ocular signal. The last one is the signal analyzed in this thesis work, which gets more attention in the field of research, especially in the aeronautical one. Therefore, this study aims to relate stress and cognitive load variations to the eye signal. The goal was achieved through initial research of the state of the art regarding the features and their trends that can be extracted from the ocular signal and related to variations in stress and MWL. Subsequently, a test session was carried out with 64 volunteer participants who were asked to perform two kinds of tests, called Stroop and N-Back tests, explicitly designed to generate feelings of stress and MWL in the subjects. The signal was acquired through the gold standard in eye-tracking Tobii Glasses 3. Finally, from the ocular signal, different types of eye movements were identified, and from them, various features were extracted. Some metrics were already analyzed in the literature, but others were never investigated; among them, a few are coming from a frequency analysis. The results showed that 83% of these features allow a binary classification between relaxed and altered states, whether stress or MWL. On the other hand, visualizing a greater granularity of trends in the two altered conditions was more difficult. In conclusion, promising results have been found in this thesis study that push us to continue research in this field to achieve the development of increasingly reliable and safe HMI systems.