Physiological stress detection using neurocognitive games

Stress has been defined as a reaction from a calm state to an excited state in order to preserve the integrity of the organism. This reaction is given by changes and pressures which provoke physical and physiological responses. Within description, stress occur in order to preserve the individuals integrity, and could appear in human signals such as heart activity, sweating or pupil dilatation. However, the presence of stress has been associated with a decrease in performance. Continuous daily stress can have negative effects on individuals’ physical and mental well-being, being linked to chronic health risks, such as cardiovascular disease, hypertension, and coronary diseases. The main goal of this Bachelor Thesis is to provide a system that determines different levels of stress using physiological signals. To achieve this goal, an experiment has been developed to induce stress to participants while their physiological signals were collected using wearable sensors. The experiment consists on a set of cognitive tasks developed by Lumosity Labs that elicits different anxiety scenarios: attention, memory, flexibility, resolve tasking and speed. Three physiological parameters have been acquired aiming to classify different stress levels: heart rate, galvanic skin response and cortisol hormone. Three devices were used to measure physiological signals; Q sensor, Garmin forerunner 310XT and Microsoft Band 2, along with a cortisol test. Finally, the classifier of stress reactions using physiological signals has been developed based on a linear regression. This complete method take into account four factors: individual performance, tasks that is been involved, chronic stress and real-time reactions. Results shows a individual scheme for different abilities played on the experiment. A characterization of Yerkes & Dodson curve is performed to evaluate his personal stress abilities using a linear normalization. Besides, features of electrodermal activity (area under the curve) and cardiac activity (ratio between low and high frequency) are selected as the ones that have more influence in terms of stress reactions. The design obtains an individual performance characterization and determine the feature more relevant on the stress detection achieving the target expectations.Ingeniería Biomédic