Real-Time Pose Estimation for Human-Robot Interaction

Text-based communication is one of the most popular categories used. Emails, SMS messages while the voice call is better than text. it connects several remote people at once and it is more personal and easier than text. currently, the video call is the best method for communication between people. After transmitting of text, voice, and video successfully the next step is motion transmission. Motion-based communication requires detection of human motion in the sending site and motivation of robot's servo motors in the receiving site. The most important step in motion-based communication is the detection of human pose in real-time and good accuracy because this detection will be driven to a servo motor. If there is some wrong in detection, it will be dangerous because the robot may be damage itself or the motion may be harmful. In this paper, we propose a realtime method for motion transmission by using deep neural networks (DNN) for the detection of human body joints, Then convert the obtained results into commands suitable to work in robot environment. And implement the results on robot so that the human motion corresponds to that of the robot in real-time.