Stress is not just stress - different stress types have different impact on well-being. Understanding one's stress and recovery patterns would be helpful in noticing and avoiding the risk for chronic stress. Currently, tools can detect stress and non-stress quite reliably, but not the type of stress. Using the stress hormone cortisol as one element in detection would help, but currently cortisol samples must be analyzed by experts. However, VTT has developed plaster-like sensors for sweat cortisol, which has potential for self-measurements. We will add this easy, near-real-time cortisol meter into the set of sensors and significantly improve the stress type detection accuracy. We will build the machine learning classifier based on laboratory experiments in which stress is induced a controlled way, and then optimize it using longer-term self-monitoring data. The study will produce novel information of stress dynamics and background factors, and an open data set for other researchers.

2022

2026