Joint Attention Detection Using First-Person Points-of-View Video

Both verbal and nonverbal communication is important to facilitate collaborative activities by multiple people. This study focused on eye gaze as a form of nonverbal communication. The proposed method consists of a local feature-matching method, an object detection method, and a combination of the two. A simulated scene was designed and filmed to evaluate the proposed method quantitatively. Experimental results show that the proposed method can achieve high accuracy when applied to local feature-matching. Furthermore, the effectiveness of the proposed method was confirmed by shooting a real scene of a medical procedure in a hospital.