Developing Standards for a Model Reference Dataset for Multimodal Human Stress Identification

Abstract

Stress is essential for improving performance and enabling day-to-day efficient functioning. But when stress becomes more than a person can handle, it can negatively impact their productivity and health. An automated system for measuring stress can assist users in keeping their stress levels within healthy bounds by tracking stress levels via physiological and physical factors. It is crucial to create and evaluate different algorithms using a reference dataset that includes multimodal stress responses in order to establish such a system. Specific standards based on findings from both clinical and empirical research must be met by this dataset. In order to facilitate the establishment of a reference dataset for multimodal human stress detection, this work presents a thorough set of criteria. The suggested specifications cover both personal and technological elements, such as the selection of sample populations, the selection of stressors, the integration of various stress response modalities, the techniques for data annotation, and the choice of data collection tools. None of the stress datasets currently in the public domain entirely satisfied these predetermined standards, according to an analysis of them. Therefore, it is essential to give top priority to creating a reference dataset that complies with these particular specifications. The comparability and reliability of study findings in the field of stress detection will be improved by such an endeavour.