Design of Natural Human-Computer Interaction for Unmanned Delivery Vehicle Based on Kinect

Abstract

There are more and more application scenarios for unmanned delivery vehicles, and traditional human-computer interaction methods can no longer meet different task scenarios and user needs. The primary purpose of this paper is to apply natural human-computer interaction technology to the field of unmanned delivery, changing the traditional mode that users can only operate unmanned delivery vehicles through the touch screen to complete tasks. The task scenarios of unmanned delivery vehicles are classified through the concept of context. Participatory design and heuristic research are used to allow users to define interactive gestures. Two groups of gesture interaction set that meet different task scenarios and can be accepted and understood by general users are designed. Based on Kinect’s deep imaging and bone tracking technology, large number of preset gesture samples are collected, and the Adaboost algorithm is used for machine training to realize gesture interaction. Through recognition and detection, it is proved that the gesture recognition achieved by this method has a high recognition rate and responding speed. Finally, based on Unity3D, the task scene of the real unmanned delivery vehicle is simulated in the virtual scene. Through the usability test of the human-computer interaction system, it is concluded that this interaction mode guarantees the task efficiency to a certain extent and improves the user experience.