Integrated gesture recognition based interface for people with upper extremity mobility impairments

Gestures are of particular interest as a HCI modality for navigation because people already use gestures habitually to indicate directions. It only takes a user to learn few customized gestures for a given navigational task, as opposed to other technologies that require changing hardware components and lengthy procedures. We propose an integrated gesture recognition based interface for people with upper extremity mobility impairments to control a service robot. The following procedure was followed to construct the suggested system. Firstly, quadriplegics ranked a set of gestures using a Borg scale. This led to a number of principles for developing a gesture lexicon. Secondly, a particle filter method was used to recognize hands and represent a generalized model for hand motion based on its temporal trajectories. Finally, a CONDENSATION method was employed to classify the hand trajectories into different classes (commands) used, in turn, to control an actuated device-a robot. A validation experiment to control a service robot to negotiate obstacles in a controlled environment was conducted and results were reported.