Stress level assessment with non-intrusive sensors

Thumbnail Image
Publication date
Defense date
Journal Title
Journal ISSN
Volume Title
Google Scholar
Research Projects
Organizational Units
Journal Issue
Stress is an involuntary reaction where the human body changes from a calm state to an excited state in order to preserve the integrity of the organism. Small amount of stress should be good to became entrepreneur and learn new ways of thinking, but continuous stress can carry an array of daily risks, such as, cardiovascular diseases, hair loss, diabetes or immune dysregulation. Recognize how, when and where it occurs has become a step in stress assessment. Stress recognition starts from 1973 until now. This disease has become a problem in recent years because has increased the number of cases, especially in workers where his/her performance decreases. Stress reactions are provoked for the Autonomous Nervous System (ANS) and one way to estimate it could be found in physiological signals. A list of a variety wearable sensor is presented to capture these reactions, trying to minimize the risk of distraction due to external factors. The aim of this work thesis is to detect stress for level assessment. A combination of different physiological signals is selected to extract stress feature an classify in a rating scale from relax to breakdown situations. This thesis proposes a new feature extraction model to understand physiological Galvanic Skin Response (GSR) reactions. Last methods conclude in incongruent results that are not interpretable. This model propose a 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. Signal processing methods of heart rhythm and hormone cortisol are included to develop a robust feature extraction method of stress reactions. A combination of electrodermal, heart and hormone analysis is presented to know in real-time the state of the individual. These features have been selected because the acquisition is non-intrusive avoiding other factor such as distractions. This thesis is application-focused and highly multidisciplinary. A complete feature extraction model is presented including the new electrodermal model named and usual heart rhythm techniques. Three experiments were evaluated: a) a feature selection model using neurocognitive games, b) a stress classifier in time during public talks, and c) a real-time stress assessment classifier in a five-star rating scale. This thesis improve stress detection overcoming a system to capture physiological responses, analyze and conclude a stress assessment decision. We discussed past state of the art and propose a new method of feature extraction using signal processing improvements. Three different scenarios were evaluated to confirm the achievement of aims proposed.
Mención Internacional en el título de doctor
Stress level prediction, Biometry, Sensors, Signal processing, Algorithms
Bibliographic citation