Predicting gaze direction from head pose yaw and pitch

Abstract - Socially assistive robots (SARs) must be able to interpret non-verbal communication from a human. A person’s gaze direction informs the observer where the visual attention is directed to. Therefore it is useful if a robot can interpret the gaze direction, so that it can assess whether a person is looking at it or some object in the environment. Gazing is a combination of head and eye movement, but detecting eye orientation from a distance is difficult in real life environments. Instead a robot can measure the head pose and infer the gaze direction. In this paper, we show that both the yaw and pitch of a human’s gaze can be inferred from the measured yaw and pitch of the human’s head pose with simple linear equations