Chronic stress leads to poor well-being, and it has effects on life quality and health. Societymay have significant benefits from an automatic daily life stress detection system using unobtrusivewearable devices using physiological signals. However, the performance of these systems is notsufficiently accurate when they are used in unrestricted daily life compared to the systems testedin controlled real-life and laboratory conditions. To test our stress level detection system thatpreprocesses noisy physiological signals, extracts features, and applies machine learning classificationtechniques, we used a laboratory experiment and ecological momentary assessment based datacollection with smartwatches in daily life. We investigated the effect of different labeling techniquesand different training and test environments. In the laboratory environments, we had more controlledsituations, and we could validate the perceived stress from self-reports. When machine learningmodels were trained in the laboratory instead of training them with the data coming from daily life,the accuracy of the system when tested in daily life improved significantly. The subjectivity effectcoming from the self-reports in daily life could be eliminated. Our system obtained higher stresslevel detection accuracy results compared to most of the previous daily life studies.
Keywords: machine learning; physiological signal processing; smart band; stress recognition.