A Smart Walking Aid for the Visually Impaired: A Case Study of the University of Ghana

Abstract

Mobility is one of the biggest challenges of the visually impaired and recent studies focusing on how technology can be used to mitigate the effects of this challenge have resulted in the development of smart canes. Current innovations of smart canes include sensors that detect obstacles as well as voice-controlled navigational capabilities. These innovations do not necessarily inform the user of what the potential obstacle is and also require an internet connection for navigation. This paper describes an offline smart cane, called WalkMaTE, with obstacle detection and classification, as well as voice, controlled navigational capabilities. The system is equipped with a pair of ultrasonic sensors placed strategically to detect branches above and obstacles below. Also, an infrared sensor is connected to a vibrating motor which detects nearby gutters and alerts the user(s). The system takes images of the user’s surroundings using a Raspberry Pi camera and determines if there are any potential obstacles in the captured image using TensorFlow (an open-source library). To have a sense of where the user is at any point in time, we use a Global Positioning System receiver module to get the coordinates which are then inputted into the navigational system. The destination of the user is provided using speech through a Bluetooth headset with the help of the CMUSphinx speech library. Routing is done with an open-source navigational system called NAVIT. The uniqueness of WalkMaTE is that it works without the use of the internet which is helpful to the visually impaired living in developing countries where the internet is not readily available. The prototype of the smart cane helps the user to get to unknown places on the University of Ghana campus and aids with the identification of obstacles in the path of the visually impaired.