This dissertation clarify how eye behaviors using an android robot with human-like appearance affects the human impression on Human-Robot Interaction. In particular, this work focuses on eye behaviors of a humanoid robot toward a human (i.e., face-to-face situation), which are an important element in nonverbal communication.
Precise control of android's eye behaviors is the most fundamental for the clarification of eye behaviors' effects. First, this dissertation proposes a gaze calibration method for an android robot, which relates an eye direction with an input control command in order to accurately adjust the android robot's gaze. Moreover, a method to evaluate the effectiveness of the gaze calibration is proposed. In the evaluation, the subjects' perception of the calibrated android robot's gaze with a human gaze is compared.
Next, this dissertation evaluates the human impression of eye behaviors using an android robot while talking to a human by comparing the motion generated by imitation and rule-based approaches. In the experiments, the subjective impression is evaluated by comparing the eye behaviors with an android robot, where the level of imitation is changed in the imitation-based behaviors and eye-contact duration and timing of eyeblinks are controlled in the rule-based behaviors.
Finally, this dissertation evaluates the human impression of eye behaviors using an android robot while listening to a human. Here, the hypothesis based on previous research is that if an android robot in a listener role imitates the human speaker's eyeblink and nodding, the android robot can display more natural nonverbal behaviors of attentive listening. Hence, this dissertation proposes a method to imitate a human's eyeblink and nodding using an android robot in real time. Then, this works evaluates the subjective impression of the imitation of the eyeblink and nodding by comparing the imitation to the rule-based eyeblink and nodding, which are simple duration-based motions, and combined motions.