Hypoxia Detection for Confined-Space Workers: Photoplethysmography and Machine-Learning Techniques

• Published in: SN computer science Vol. 3; no. 4; p. 290
• Main Authors: Wei, Yixuan, Jin, Longzhe, Wang, Shu, Xu, Yifei, Ding, Tianqi
• Format: Journal Article
• Language: English
• Published: Singapore Springer Nature Singapore 01.07.2022; Springer Nature B.V
• Subjects: Accident investigations; Algorithms; Bioelectricity; Blood pressure; Computer Imaging; Computer Science; Computer Systems Organization and Communication Networks; Confined spaces; Cycle time; Data Structures and Information Theory; Decomposition; Experiments; Fatalities; Gas detectors; Gases; Hemoglobin; Hypoxia; Information Systems and Communication Service; Internet of Things; Machine learning; Measurement techniques; Medical equipment; Monitoring; Original Research; Oxygen; Parameters; Pattern Recognition and Graphics; Software Engineering/Programming and Operating Systems; Vision; Wavelet transforms; Wearable computers; Wearable technology; Workplaces; China; Hypoxia; Internet of Things (IoT); Machine-learning algorithm; Photoplethysmography (PPG); Confined space
• ISSN: 2661-8907; 2662-995X; 2661-8907
• DOI: 10.1007/s42979-022-01162-5

Oxygen-deficiency is a cause of fatalities in confined-space workplaces. Most research and projects have been conducted to reduce work-related accidents by external measurements, while works addressing the early warning of workers’ hypoxia state through a bio-electrical signal have rarely been conducted. In this paper, we present a hypoxia detection system based on non-invasive photoplethysmograph (PPG) measurement using machine-learning (ML) algorithms. The PPG signals obtained from 22 subjects underwent preprocessing and features extraction steps. Time-domain features, rising and falling slopes, and amplitude parameters were adopted to train and test the ML algorithms. In addition, an Internet of Things- (IoT) based smart wearable device and monitoring system were developed to measure the vital parameters of workers in confined spaces. Cardiac cycle time of all the participants except subject 8 decreased significantly ( P  < 0.01) throughout the oxygen-deficiency trial. The PPG signal complexity essentially decreased ( P  = 0.006) when the concentration of environmental oxygen declined. The DT algorithm with nine input parameters outperformed the other algorithms in prediction accuracy (Acc = 0.9). The results show that the features extracted from the PPG signal can be adopted as important indicators to revealing the hypoxia state of workers. The device and system automatically measure and analyze the PPG signal with the location details of caregivers and notify the monitoring staff in the case of an emergency.