Reliable stress measurement using face temperature variation with a thermal camera in human-robot interaction

For companion robots to properly interact with people, they should be able to gain awareness of their effect on the person's state and learn to behave accordingly. Advances in physiological monitoring sensors, such as infrared cameras, may enable robots to get an insight on a person's emotional state, which has a strong correlation with the physiological parameters. Performing this monitoring in a contact-free, user transparent manner is an important requirement. In this paper, we used a thermal imaging sensor mounted on the Meka humanoid robot, to remotely measure temperature variations on the face of the person while interacting with it. The scenario was a card-based quiz with the robot, which was aimed to induce a stressful state and 16 participants took part in the experiment. Several conditions with variations in the distance, height, and gaze of the robot were tested. A statistically significant variation in temperature in the nasal area of the face was measured while the robot was standing in the personal space of the participant. The first preliminary results encourage us to believe that temperature variations across various areas of people's faces can hint companion robots to adjust their behavior in human interactions.