Real-time Face Tracking for Human-Robot Interaction

This paper proposes a real-time human face tracking to be used for human robot interaction (HRI). The system consists of human face like region detection, its tracking and camera motion control according to the displacement between the time interval. The former task is performed via utilizing Viola-Jones algorithm as the face tracking uses the Kanade-Lucas-Tomasi (KLT) algorithm with various poses conditions. The camera is controlled by the displacement according to its displacement between image frames as a tracking result from the previous stage. The experimental results show that human faces were successfully tracked even while the users are wearing glasses, hat, or lateral face postures in real-time. The experiments were done in the 28.32 frames per second as maximum speed.