Skin color detection based on various chrominance pairs from multiple color spaces for face detection in Human-Robot Interaction (HRI)

Abstract

A significant way for a mobile robot to detect human presence is through human face detection. Development and implementation of face detection on mobile robot remains a great challenge due to the dynamic nature of the robot and the environment. Thus, the face detector has to be fast and flexible to cope with this problem. Main objective of this paper is to present a skin color detection method for face detection using multiple color spaces. Skin color detection is implemented to find human presence based on the robot?s vision by combining chrominance information of images from multiple color spaces. Later, the skin region is further verified using crisp rules and eigenface techniques. Skin color detection serves as the backbone for face detection. The face detector is implemented on a mobile robot for human-robot interaction (HRI). Embodiment of the robot enables it to interact with a human through motion reaction and a graphical user interface. A tree-based decision making technique is used to combine the results of face detection to accommodate different HRI behaviors. Experiments for face detection demonstrate the advantages of using multiple color spaces. Overall, the face detection using crisp rule has an accuracy of 86.61%.