Behavior prediction for several seconds using human body model and non-contact sensor

摘要

Based on social background such as aging and diversification of individual's lives, we propose the concept of Biofiedbuilding. The Biofied building is a concept that provides a safe, secure and comfortable space for residents by using smallrobots. The small robots used in the Biofied building sense the human position and motion. However, in the current systemit is difficult to prevent accidents such as falls. The robots are not designed not disturb residents. Therefore, in this research,in order to realize more natural space control of the robot, the purpose is to incorporate motion prediction in the system inaddition to motion recognition. Many conventional motion prediction methods use RNN (Recurrent Neural Networks),which is considered to be difficult to implement due to the necessity of big data and the large calculation cost. Furthermore,in RNN, it was a problem that physical constraints of the human body were not taken into consideration. Therefore, thisstudy aims to propose a simple motion prediction method using a human body model based on dynamics.