Gesture Based Tele-Operation
Sambit Kumar Dash
April, 1998
Advisor: Amitabha Mukherjee
MASTER-SLAVE mode: The robot following the hand Abstract
Teleoperator is traditionally done via joystick or keyboard from a dedicated control station. Gestures provide a rich intuitive communication interface for controlling real devices. Gesture recognition has been mostly based on specialised gesture pick up devices like datagloves or magnetic position sensing Polhemus. With the advent of better image capturing and processing systems these special gesture capturing devices can be replaced with nonintrusive image processing systems. This work is an attempt to integrate computer vision to the control of real devices. We develop two different gesture libraries in symbolic and pointing modes, and use them to control a mobile robot (symbolic) and a manipulator (static pointing gestures).
We observe that one of the strengths of symbolic gestures is that the same gesture set can be used both for mobile robot and a manipulator. By combining spatial directives for incremental motion with symbolic instructions for controlling velocity, gestures can provide highly accurate control.