Detection of physical stress using multispectral imaging

• Published in: Neurocomputing (Amsterdam) Vol. 329; pp. 116 - 128
• Main Authors: Hong, Kan, Liu, Xiaoling, Liu, Guodong, Chen, Wentao
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 15.02.2019
• Subjects: Multi-spectral imaging; Physical stress; StO2; Physical stress; Multi-spectral imaging; StO2
• ISSN: 0925-2312; 1872-8286
• DOI: 10.1016/j.neucom.2018.10.011

This study explored the potential of multispectral imaging for detecting physical stress on humans. Multispectral images were obtained from recruited participants, and a max completion-based differential tissue oxygen saturation (StO2) algorithm model was proposed and established without background information. Further correlation analysis showed that multi-subject facial StO2 fluctuations have similar intra-class affective patterns under physical stress. The algorithm model was verified with respect to physical stress ground truth to classify the baseline and physical stress status. The algorithm implements non-contact, non-intervention, and background-free physical stress detection. The algorithm achieved promising results in the experiment with an accurate rate of higher than 90%, which will provide an excellent foundation for future industrialization. Experimental results demonstrated that multispectral imaging, as a non-invasive method, has the potential to identify physical stress on humans.