Development of a wearable guide device based on convolutional neural network for blind or visually impaired persons

This study proposes a design for a wearable guide device for blind or visually impaired persons on the basis of video streaming and deep learning. This work mainly aims to provide supplementary assistance to white canes used by visually impaired persons and offer them increased freedom of movement and independence using the proposed wearable device. The considerable amount of environmental information provided by the device also ensures enhanced safety for its users. Computer vision in the proposed device uses an RGB camera instead of the RGBD camera commonly used in computer vision. Deep learning is applied to convert RGB images into depth images and calculate the plane for detecting indoor objects and safe walking routes. A convolutional neural network (CNN) is adopted, and its neural network structure, which is similar to that of the human brain, simulates a neural transmission mechanism similar to that triggered in human learning. Therefore, this system can learn a large number of feature routes and then generate a model from the learning result. The proposed system can help blind or visually impaired persons identify flat and safe walking routes.