Fatal accidents occur at construction sites. Incidents in- volving dangerous situations but not reaching the category of fatal accidents also take place. When an incident occurs, workers typically generate an on-site physiological reaction. In this paper, an automatic detection system is proposed to automatically identify incidents by measuring biological sig- nals related to emotions of on-site workers, such as heart rate and the masseter muscle. In the first stage of this study, some virtual reality (VR) video-based experiments were conducted with some wearable sensors to confirm whether the afore- mentioned biological signals are suitable to detect incidents. While watching the VR videos, biological signals of the sub- jects were measured using different wearable sensors. The experimental results corroborated that the proposed sensing method is suitable to detect construction-site incidents.