Human Body’s Orientation Estimation Based On Depth Image

Abstract

In human-robot interactions, robots that often interact with humans need some information from humans to be able to communicate. In human-oriented robots, the robots must pay attention wherever the human will go. Therefore, the human-oriented robots are more flexible. To estimate wherever the human will go, it can use the direction of the human or what is called intention. Intention can be taken from the perspective of the human body. An algorithm is needed to estimate the orientation of the human body. In this study, using image processing from depth image is expected to be able to estimate the orientation of the human body when a human is not moving (static orientation) so that it can be used by the robots to predict the intention direction of a human. One method that can be used to classify the orientation of a human body is the Support Vector Machine (SVM) so that it gets a classified orientation value. Then from the classified orientation value to get the static orientation value, Kalman Filter is used. Orientation classification with SVM method has accuracy of 96.528%. Static orientation estimation using this method produces precision value of 0.097 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">0</sup> .