Stress is a feeling generated to face certain challenges where human body responds by activating the nervous system and hormones. This reaction changes some physiological activations such as increasing respiratory rate, blood pressure, metabolism, muscles are putting on tension, pupil diameter decreases, among other variations. All these physical changes prepare the individual to react quickly and effectively to complete a task.
A little stress helps to be ready to overcome any confrontation, but if stress increase more and more, can be trigger in a wrong decision due to an anxiety situation. Even if stress became continuous , it supposes a decreases of performance and could be detrimental to health. Recognize how stress affects, when occurs and where it can be founded has become a step in medical health industry.
The aim of this thesis is to monitor a subject while is achieving a stressful task and make a relationship between stress reactions and the performance obtained. Quantify or scale physiological activations vary on the person, and even in the same person, are not always shown similarly. Past studies have used self-report questionnaires to relate stress and performance but are obtrusive, requiring the attention of the subject, and/or the level scale is not normalized. Another alternative way could be to monitor these signals as stress features and obtaining a objective value.
This thesis propose a system acquisition to acquire signals in non-intrusive way, keeping the quality and avoiding an array of possible artifacts or out-layers that make signal processing difficult. The activities that can be measured in this way are: heart rhythm, electrodermal and cortisol hormone.
In one hand, this thesis proposes a new feature extraction model to understand physiological electrodermal reactions. Past methods conclude with incongruent results that are not interpretable even to differentiate between relax or stress situations in a time-line. We propose a new robust algorithm that can be used in real-time (low time computability) and results are sparse in time to obtain an easily statistical and graphical interpretation.
On the other hand, this thesis presents a feature extraction method of stress reactions including signal processing methods of heart rhythm and cortisol analysis founded in the literature. Finally, features are analyzed in different states using machine learning algorithms to extract conclusions.
Three experiments are evaluated in a stress elicited environment: a) the participant is elicited stress playing a variety of neurocognitive games , b) the subject is monitored while is discussing a public talk , and c) reactions of a operator of unmanned aerial vehicles are analyzed while is simulating flight mission.
The conclusions obtained for each experiment are: a) the most relevant features obtained to differentiate between relax and different classes of games are: sparse reactions of electrodermal activity, heart rate variability and the frequency ratio in heart rhythm. b) Using only these features it can be differentiated between: pre-talk, talk and questions time. Finally, c) this thesis improve a system to capture physiological responses, analyze and conclude in real-time a stress assessment classification in a five levels rating scale.
© 2008-2024 Fundación Dialnet · Todos los derechos reservados