Navigational Aid for the Blind Using Deep Learning on Edge Device

Abstract

Navigation plays a major role in the life of visually impaired people as they cannot get around easily without the assistance of escorts and hence, they must rely on their sense      of touch and hearing in order to tackle the obstacles on their way. The use of guide dogs, sticks have not  proven  helpful.  They are not able to make  the  visually  impaired  individuals  de- pend on themselves. The major challenge faced by a visually impaired person is the difficulty in recognizing the surrounding objects and traversing from one place to another. In order to solve these problems, there are several computing techniques explored for real-time navigation for visually impaired such as Artificial Intelligence, Deep Learning, Machine Learning. With the help of deep learning algorithms, it is easy to identify different objects and these algorithms can be implemented in the form      of a mobile navigation application. Deep Learning is basically     a subdivision of Machine Learning which involves algorithms enlivened by the working structure and functioning of the human brain known as Artificial Neural Networks. It involves teaching the computer to screen the inputs through a set of layers to determine the prediction and classification of the information. Deep Learning is the representation of the working of human brain and its screening  of  information.  The  observations  are in the form of text, sound or images. The recognition of such domestic sounds, if implemented on a mobile device, can lead     to new and important applications, one of those is navigation.  The deep learning algorithms are used for acknowledging and perceiving objects and further applied as mobile navigation system. In this project, with the help of deep learning techniques we were able to implement blind navigation system for visually impaired individuals in order to assist them with the help of a mobile app which makes use of a smartphone camera and is able to detect the obstacles on the way and informs blind people about those obstacles through audio signals which are formed by the gTTS library used for conversion of text to speech and hence,  can be heard with the help of earphones which further helps the visually impaired individuals to move further in- dependently without colliding with the obstacles.