Real-time Vision-based Gesture Recognition for Human Robot Interaction

Real-time gesture recognition is mandatory for gesture based human-robot natural interaction. Several glove-based approaches are introduced for this purpose, but they are un-natural and force user to wear some cumbersome glove. This paper presents a vision-based gesture recognition method for human-robot interface. Hand poses and face are detected using multiple feature-based template matching techniques from the three largest skin-like regions that are segmented using skin color information from YIQ color space. Gesture is recognized using rule-based approach whenever the three skin-like regions at a particular image frame match with the predefined gesture. Gesture command is sent to robot through a knowledge based software platform for human-robot interface. In this paper, a method has also been proposed to detect left hand and right hand relative to face position as well as, to detect and locate its position. The effectiveness of the proposed method has been demonstrated over the interaction with a pet robot named AIBO