Inferring workplace safety hazards from the spatial patterns of workers' wearable data

Hazard identification in construction typically requires safety managers to manually inspect an area. However, current approach is very limited due to the dynamic nature of construction sites and the subjective nature of human perception. Using wearable inertial measurement units (WIMU), previous literatures revealed the relationship between a worker's abnormal gait patterns and the existence of slip, trip and fall (STF) hazards. Though the prior work demonstrated the strong correlation between STF hazards and abnormal gait patterns, automated hazard identification is a challenging issue due to the lack of knowledge on decision threshold on identifying hazards under different construction environments. To fill the research gap, this study developed an approach that can automatically identify the STF hazards without knowledge about thresholds by investigating the spatial associations of workers' abnormal gait occurrences. An experiment simulating a brick installation was performed with different types of STF hazards (e.g., poor housekeeping), and results demonstrate the feasibility of STF hazards identification with the developed approach. The results highlight the opportunities of revealing potential accident hotspots via an efficient and semi-automated methodology, which overcomes many of the limitations in current practice.