A human-robot interaction interface for mobile and stationary robots based on real-time 3D human body and hand-finger pose estimation

A real-time human-robot interaction framework with robust background invariant hand gesture detection

“…Such techniques have recently been outperformed by modern deep learning ones like convolutional neural networks [38]. The authors of [39] propose a human-robot interaction system for the navigation of a mobile robot using Kinect V1. The point cloud acquired from Kinect V1 is fit on a skeleton topology with multiple nodes, to extract the human operator pose.…”

“…In [60], human body localization is performed using laser sensors, and its sub-parts are obtained through Kinect with the OpenNI library as in [61]. In [39], the authors localize the human body, inspired by [62], by merging clusters of the point cloud obtained from the Kinect V1 after voxel filtering and ground plane removal.…”

Call to Redefine Human-Robot Interaction: From Social Robots to Gun Shooting Humanoids

“…However, close proximity would not be sufficient in redefining the interaction of today. For example, there are nano-robots taking on certain assignments in some patient's bloodstream to repair a damaged cell back to health (Felfoul et al, 2016). In these cases, an intact state is observed rather than some proximity.…”

“…Perhaps the discussion would not be complete by including neither the appearance nor the movement, but also the learning and creating abilities. Robots ability to learn quicker than an average human (Rossi & Lee, in press), even the gestures (Ehlers & Brama, 2016) and the social cues by often mimicking the Human-Human-Interaction (Wehle, Weidemann, & Boblan, 2017), needs to be accounted, too.…”

A Framework for Real-Time Physical Human-Robot Interaction using Hand Gestures

“…The OpenNI and NITE middleware are used to extract skeleton information of the human user. Authors in [5] and [6] also propose HRI scenarios using Kinect. Mostly the researchers have used OpenNI or Microsoft SDK to extract the human-skeleton, which is a model based skeleton tracker having several discrepancies including the need of initialization pose, not being able to detect gestures on which the model is not trained on, and noisy detections.…”