User Tools

Site Tools


Action unknown: copypageplugin__copy
projects:year9:9a.008.ul

9a.008.UL - Privacy-aware Stress & Activity Recognition Using Wearables in Hospitals

Project - Summary

Automated methods to detect human behavior enables the intelligent systems to better adapt to human behavior. Recognizing human activity and emotion in buildings is import for various reasons such as assessment of employee well being, energy and space optimization, etc. Many organizations such as hospitals analyze stress using survey based methods. But these methods are cumbersome, and are subject to recall bais, noise and uncertainity. The primary goal of this project is to improve the performance of stress detection methods with wearables in real-world conditions. Studying stress “in the wild” in a work environment is complex due to the confluence of many social, cultural, and individual factors in dealing with stressful conditions. The project is an extension of previous years work on stress detection. We designed a study to compare the performance of stress detection with wearables and a traditional survey for nurses in a major hospital. We leveraged real-time stress detection methods and tools developed from prior work. The project's key outcomes are improved stress detection models, a human subject stress study on nurses in a major hospital, a novel stress detection data-set, and detection of basic activities of nurses from accelerometer sensors.

Project - Team

Team Member Role Email Phone Number Academic Site/IAB
Raju Gottumukkala PI raju@louisiana.edu (337) 482-0632 UL Lafayette
Christoph Borst Co-PI Christoph.Borst@louisiana.edu (337) 482-1023 UL Lafayette
Ziad Ashkar Co-PI ziad.ashkar@louisiana.edu 337-482-0609 UL Lafayette
Satya Katragadda Co-PI satya@louisiana.edu (337) 482-0625 UL Lafayette
Funded by: TietoEVRY

Project - Novelty of Approach

  • Existing stress detection methods use more than one wearable to detect stress and the stress detection data-sets are in controlled laboratory conditions. However, our model uses only a wristband with minimum intrusion for the subject during their daily work.
  • Investigation of between-subject stress detection is challenging due to personalized stress detection methods. In this study, we investigated the between-subject stress detection to improve
  • It provides an “in-the-wild” data-set that contains more than 1200 hours worth of data for stress detection of the nurses. This study has been done in the COVID-19 pandemic and can be a valuable data-set to investigate the pandemic effect on the nurses' stress in the hospital.
  • The activity data-set is collected more than 12 hours of ten participants performing different activities, namely computer work, sitting, walking, running, ascending, and descending stairs. This data set used only a wristband to detect different activities, and it showed promising results in terms of accuracy.

Project - Deliverables

Deliverables
1 Methods to anonymize user’s location in a building when using wearables and indoor-presence sensors
2 Software & algorithms for stress & happiness detection

Project - Benefits to IAB

We believe this study can be helpful for researchers in many domains. First, this data set is useful for researchers to improve the stress detection performance of the models. Second, we provided accelerometer data and the stress that helps detect activities to understand the relationship between activity and stress. Finally, we provided the stress survey results in stressful events that help the researchers in several eras, e.g., human resources, human factors, and organizational psychology, to associate between biometric signals and stress-related factors during the COVID- 19 outbreak. Even without the signal data, the data-set would be additionally valuable to understand the differential distribution of various work-related stressors during the pandemic.

Moreover, the ongoing study on multi-modal representation fuses different signals to improve stress and emotion detection algorithms. In this study we started a new study and data collection to find the relation between stress and facial features.

Project - Documents

projects/year9/9a.008.ul.txt · Last modified: 2021/08/24 09:28 by sally.johnson