Indoor Objects Detection and Recognition for Mobility Assistance of Visually Impaired People with Smart Application

Abstract

Visually impaired people are individuals who have a partial or complete loss of vision. This condition can vary in severity, from mild to profound, and can be caused by a variety of factors, including genetics, injury, illness, or aging. It's important to note that visually impaired people are a diverse group with different needs and abilities, and they should be treated with respect and given equal access to opportunities and resources. Smart applications can be extremely helpful for visually impaired people by providing them with information and assistance in navigating their environment. Moreover, can greatly enhance the independence and quality of life for visually impaired people. Many applications have been proposed with individual features set and other concerns due to are expensive, difficult to use, less affordable, less accessible and are an overhead while travelling. To solving these issues, in this paper we design an android Application - Smart Vision (SV) to aid the Visually Impaired people in every possible realm of daily pursuits like for their diurnal regressions around the house/office place, taking notes of their daily affairs, recognizing the people around, identifying the colours etc. Smart Vision would enable the visually impaired to have a better user experience as the whole working of the application is divided into 6 modules - Obstacle Avoidance and Navigation Module, Digital Assistant Module, Scene Description Module, Light Detection Module, Colour Detection Module and Face Detection Module, which can be easily selected by giving a single voice command. Smart Vision is exclusively designed by keeping the memory and battery constraints of the application within permissible limits ensuring the reliability, portability and cost effectiveness to the grieved end user to make life beautiful for them. Smart Vision enables detection of the desired object with a 90% accuracy, face recognition with an accuracy of 87%, colour detection with an approximate accuracy of 75%, Digital assistant module accuracy as 91% and light detection (for light and dark intensities) minimum accuracy is 82%. It has been interpreted that the Average Response Time is under 3 seconds which makes it a high-speed device because the Raspberry pi is used for connecting the components which reduces the transit time. Finally, a comparative study has been done with the existing topologies and it is so found that the Smart Vision is applicable for both indoor and outdoor environments.