Cognitive assistance for the visually impaired using haptics presentation of environmental information

In this work-in-progress (WIS) paper, we propose an approach to assist the visually impaired in walking independently by presenting human flow and sidewalk edge information. To achieve this, a robust estimation of the direction of human flow and a method for detecting the edges of sidewalks are presented. Image learning for sidewalk estimation is performed using binary semantic segmentation that classifies sidewalks and other areas to detect sidewalks regardless of pedestrians and shadows on sidewalks. The effectiveness of the methods was evaluated.