A study on an assistance by mobile robots in preventing the elderly from falling

Abstract

In this paper we reinvestigate an increase tendency of recently elderly’s accidents and the place “Entrance” having high probability of accidents. Further we try to apply DCNN (Deep Convolutional Neural Network) to recognize environment and to analyze risk of fall that is a main factor of accidents for the elderly people. This study proposes a 3-stream DCNN method to detect changing factors and to assist elderly by mobile robots. At the point of view about images, first stream input is the current environmental image, second stream input is the visualized optical flow, and third stream input is the changing rate of visualized optical flow. For an application example, the method detects the environmental sound, controls camera or robot direction to sound sources, detects liquid on the floor to the entrance, estimates the severity of accidents, and makes a warning for users. The experimental results show that our method is functioning and useful. Our approach is able to use as the means to detect changing factors which are the trigger of accidents.