Haar-Feature Based Gesture Detection of Hand-Raising for Mobile Robot in HRI Environments

This paper proposes a method for hand-raising gesture detection that can be used for mobile robots in indoor environments. Different from traditional methods which are capable of the detection of handraising gesture with a static camera, our method can process the detection on a non-stationary platform. At first, haar-like features of raised arms are extracted and a cascade adaboost classifier is trained. Then the classifier is used to detect whether there are hands raised in specific regions which are established by results of face detection. The detector completes the detection in different scales at all locations of an input image. Experiments on a mobile robot are implemented in indoor environments where several persons are walking or standing randomly. Experimental results show that our method is suitable for realtime hand-raising gesture detection in Human-Robot Interaction in indoor environments.