Sensing stress: stress detection from physiological variables in controlled and uncontrolled conditions

Recently, general concern about work-related stress is increasing. Chronic stress induces a number of mental and physical health problems that impact personal life, organizations and society. Timely detection and reduction of stress could prevent these health problems and their negative effects. Stress causes rapid activation of the autonomic nervous system and this activation can be measured in a number of physiological variables. The goal of this thesis is: to assess the feasibility of constructing personal models for the relation between mental stress and physiological variables, for use in ambulatory stress management systems. Four studies were performed in which physiological variable were measured, as well as self-reported stress measures and context variables. Stress induced reactions in the physiological variables, but the pattern of the reactions varies from person to person. The main conclusions of this thesis are that using physiological variables for mental stress detection is feasible, personalization is necessary due to large variations among persons, and that ambulatory measurements are feasible if an unobtrusive and low-power sensor is available. The most common features used in stress estimation are blood pressure, heart rate and skin conductance. Other features such as heart rate variability, EMG and temperature are relevant for some subjects, but not for others. Respiration rate could be a useful feature but is heavily influenced by speech.The main difficulty in the research field that needs attention in future work is that there is no recognized reference measure available that is known to resemble actual mental stress level accurately.