links<! -- ========================== GROUP PEOPLE ========================== ->
<! -- ========================== GROUP PAGES/TABS ========================== ->
Stress Monitoring - overview<! -- ========================== PAGE CONTENT ========================== ->
Early and reliable detection of stress helps medical professionals, and coaches to better manage team performance and health. Differentiating external biogenic or physical stresses from psychological stresses is even more relevant to improve performance under extreme circumstances. Stress directly stimulate the sympathetic nervous system that primarily increases heart rate and a host of other responses like salivary secretion, electro-dermal activity, muscle tension, and brain activity. However, most of these methods are not suitable for real-time stress monitoring because they are invasive, expensive, impede regular functioning of subjects, and are time consuming. Current ways of analyzing ECG signals cannot distinguish reactions to external or internal stressors, additionally optical PPG is inaccurate when recorded during intense physical activities. Most current wearable monitoring systems are organized as a vertical silo with the HRV analysis being performed in a non-real-time fashion in the cloud with HRV results being sent to the user periodically. No device can show more complex parameters like HRV or total stress in a real-time fashion. Wearable monitoring becomes difficult in situations where wide bandwidth communication is impossible.
Stress detection has a huge potential for disease prevention and management, and to improve the quality of life of people. Also, work safety can be improved if stress is timely and reliably detected. The availability of low-cost consumer wearable devices that monitor vital-signs, gives access to stress detection schemes. With DeStress demonstrate a stress monitoring system based on machine learning that learns based on time-based and frequency-based data what is a stressed person and what is a relaxed person. We have shown that stress can be measured with a 1-minute time resolution.
We can detect stress with improved accuracy and then accurately assess whether the stress is due to emotional or physical stress / temperature stress. The main objective was to quantify the ability of a person to make critical decision and to assess and predict the reduction of this ability by the different intensities and categories of stress. Our main contribution is an edge(hybrid) system that can perform inference at the edge and thus overcome computation and connectivity limitations. In these situations, the analysis needs to be performed local with only a simple result being sent periodically to the base station. Firemen need to enter into extreme environments where it is not possible to have a high bandwidth communication to the outside. For this reason, a localized analysis of the stress is essential - an ideal situation for an intelligent hub or edge system. For occupations such as firefighting that are prone to acute stress, detecting and managing mental stress helps to save lives. We will continue to test the system with first responder organizations in our region and later in several European Countries.
Heart rate variability (HRV) was derived from wearable device data to reliably determine stress-levels. In order to build and train a deployable stress-detector, we collected labeled HRV data in controlled environments, where subjects were exposed to physical, psychological and combined stress. We then applied machine learning to separate and identify the different stress types and understand the relationship with HRV data. The resulting C5 decision tree model is capable of identifying the stress type with 88% accuracy, in a 1-minute time window. We created an integrated system to acquire expert labels in real-time from firefighters during their training in a Rescue Maze. A model to detect and classify physical and mental stress in real-time using HRV analysis was developed using low cost wearable technology and deployed without impeding the person’s activity/mobility. The model was trained using HRV data collected from participants in laboratory conditions. During this training the best performing model, based on C5 decision tree, showed a precision, recall and F-score of close to 90%. When applied to data from a firefighter smoke diving training exercise, the models predicted the correct stress classes and distinguished mental stress from physical stress.
- Pluntke, S. Gerke, A. Sridhar, J. Weiss and B. Michel, “Evaluation and classification of Physical and Psychological Stress in Firefighters using Heart Rate Variability”, EMBC Conference, July 24-27, 2019, Berlin, Germany.