Imitation of Human Arm and Hand Natural Movement

In this research the imitation of the human arm gesture is attempted for incorporating its application to telecommunication. A unilateral system to achieve this is implemented using robotic arm as the actuator for reconstructing arm gestures, and computer vision for the sensor. Two cameras are used for tracking the motion of the shoulder, elbow and hand from both the front and side view of the user in order to calculate the joint angles that represent the arm pose. The tracking is done using grayscale correlation. Issues such as arm motion tracking, occlusion, arm motion transmission delay, imitation accuracy, actuator size, performance, acceptance and appropriateness are investigated. The results from the subjective evaluation show that imitation accuracy is not really a paramount aspect in the reconstruction of the gestures, which is counter-intuitive. The low bit rate nature of gesture transmission makes the handling of its traffic attractive even in the network such as the Internet. A maximum allowable delay that should be maintained for having in imperceptible delay in the gesture reconstruction is proposed based on the result of the subjective evaluation.