STRESS ASSESSMENT THROUGH GALVANIC SKIN RESPONSE USING SMART WATCHES

In recent years, many studies regarding the analysis of stress have attempted to replicate in the laboratory stress conditions that might occur in everyday life. Wearable devices were employed to collect and evaluate physiological signals in order to do this. Thus, the purpose of this study was to assess a stress condition by analyzing the skin conductance response signal through the use of the Empatica E4 wearable device. To understand the origin of the physiological signal analyzed, it was important to first understand the functions of the sweat glands and how sweating affects the electrodermal activity (EDA). EDA was divided into tonic component, slow and constant, and phasic component, more rapid and reactive, to then extract the characteristics and determine a stress situation. The latter can stress the body over time and have negative effects on health. The most common signs of stress are mood changes, damp or sweaty palms, difficulty sleeping and headaches. The body reacts to stress with a so-called "fight or flight" response, during which certain hormones, such as adrenaline and cortisol, are released. In accordance with the literature, the most commonly used way to detect stress is through the use of wearable devices such as smart watches as they are convenient and unobtrusive. Therefore, since this study is primarily aimed at detecting stress in the work environment, Empatica E4 was chosen for the above reasons. EDA reflects the activity of the sympathetic sudomotor nerve and is related to the electrical conductance of the skin, which varies with sweat production. EDA increases in response to a stress stimulus so that the skin conductance response (SCR) has a rapid increase that peaks in approximately 1 second, followed by a temporal decay with a half-life of approximately 3 seconds. The experiment was conducted on eight adult subjects for a duration of approximately 30 minutes for each participant. All wore the Empatica E4 bracelet on their nondominant hand for the duration of the test, making sure to hold it steady. The test consisted of a relaxation period followed by a moderate stress period, in which the subject had to think of a speech that was then to be presented in front of the examiner, and a more intense stress period consisting of an arithmetic test followed by a final relaxation phase. After the experiment the EDA signal was extracted which was then analyzed in Ledalab, a Matlab-based software, using Continuous Decomposition Analysis through which the tonic and phasic components of the signal were extracted. Subsequently, an algorithm was implemented in Matlab which allowed the signal to be divided into smaller windows of 60 seconds and the most important features such as the number of peaks, the average amplitude of the peaks, the standard deviation and the maximum value of the peaks were extracted. After conducting a statistical analysis on the number of peaks in each phase of the experiment, it was concluded that during the most intense stress phase there were more peaks with higher amplitude. In fact, in almost all subjects the number of peaks during the first half of the experiment was zero while during the second half, that means from when the oral exposition of the speech followed by the arithmetic test began, the subjects showed an average of about 20 peaks per minute. We note, however, differences between more anxious subjects, who reached even 40 peaks per minute, and less anxious subjects, with an average of about 15 peaks per minute. This confirmed that through the use of smartwatches, specifically Empatica E4, it is possible to detect and assess a stress condition through the analysis of the electrodermal signal of the skin.