Assistive technological solutions for visually impaired

Abstract

According to World Health Organization s report, one­sixth of the world population is suffering newlinefrom vision impairment. In the past decades, many efforts have been done in developing several newlinedevices/solutions to provide support to the visually impaired (VI) and enhance the quality of newlinetheir lives by making them capable to lead normal life. Many of those devices are either heavy newlineor costly for general purposes. It would be exceedingly unfair to leave the VI off in these day newlineof technological developments, where humans are exploring it in every field. As the figures newlineof VI are increasing, the necessity of having a solution for navigation and orientation has also newlineincreased. newlineThe main focus of this work is to develop solutions to help VI. First of all, a survey has been newlinedone to know about all the existing solutions developed for them. A brief comparison has been newlinedone between all those solutions based on the various evaluation parameters to understand their newlinelimitations. Once the survey is done, a hardware integrated software solution in the form of cane newlinehas been developed to assist them in the general day­to­day activities such as object detection, newlinetext detection, and path texture detection. The cane has been shaped using edge device, Nvidia newlineJetson NANO/raspberry pi, a 3D camera, and other sensory devices attached to it. For the initial newlinetesting, existing detection models have been used and deployed on it. To understand the memory newlineconstraints and the compatibility of different models on edge devices, experiments have been newlineperformed. This helped in knowing how the edge devices performed when a certain model, newlineeither lightweight or heavy, was deployed on them in terms of detection time and accuracy. The newlinecustom­developed deep learning models have also been deployed on the edge device attached newlineto the cane and then the performance analysis has been performed. newline newline