UC San Diego

Motivated by the Electronic Health Record (EHR) system's demand of capturing patients' multimodal activities and the wide application of gaze detection, we develop a minimally intrusive gaze detection system with Microsoft Kinect sensor and test its performance in a simulated clinical environment. Traditional methods require either a close distance between the camera and the user or a fixed head pose which may severely interrupt the clinical workflow and the interaction between the physician and the patient. Compared with the traditional methods, our system allows a wider range of detection, while achieving an accuracy around 70%