Development of a Walking Assistance Navigation System for the Visually Impaired that Utilizes Walking Trajectories

Abstract

The current challenge in real-time path generation for visually impaired individuals' walking assistance using image processing lies in the lack of temporal information,which prevents the generation of accurate paths considering social interactions with surrounding people.By utilizing the walking trajectories of sighted individuals as the ground truth data for a neural network,it becomes possible to generate safe paths resembling those of able-bodied individuals.Additionally,by combining three processing methods,namely Convolutional Neural Network (CNN) for image processing,Principal Component Analysis (PCA),and Gated Recurrent Unit (GRU) for temporal information processing, visually impaired individuals can select safe and accurate paths while considering social interactions with people around them.