Human Direction Determination for Visually Impaired Support Indoor Mobile Robot

Abstract

Independent mobility is an essential part of the daily life of many visually impaired individuals. To assist them in navigation and improve their quality of life, various designs have been developed, including portable devices such as electronic canes, navigation suitcases, and guided robotic dogs. However, when it comes to indoor environments, constantly carrying an object alongside is not practical. Therefore, there have been numerous studies on mobile robot systems for navigation and assistance in indoor environments. One of the requirements for such a non-contact robot system is to determine the direction of the visually impaired individual to provide accurate assistance. This article focuses on developing algorithms for determining the direction of the object that needs assistance in indoor environments. The object's information is collected using a 3D camera, and then techniques such as object tracking, skeleton framework construction, and Gaussian filtering of depth images are used to calculate the object's direction angle. The proposed algorithm was tested with a representative sample of visually impaired individuals in a laboratory setting. The results showed high accuracy for static objects and acceptable performance for dynamic objects. This forms the basis for applying this algorithm to mobile robot systems that assist visually impaired individuals.