Development of a mobile object recognition system to assist the visually impaired

Abstract

The popularization and easy use of mobile devices (such as smartphones, tablets, among others) made possible to share photos and data anytime, anywhere. In this context, this project consists in exploring such devices presenting a mobile system (application) to assist the visually and hearing impaired, with the objective of identifying and locating objects and providing information about the user favoring the independence of their mobility. For this, virtual reality will be used and increased as a support for information presentation through three-dimensional gestures (language), sign images and sound. In addition, hybrid image descriptor will be developed for the recognition of objects in multimedia data generated from mobile devices. Therefore, a visual and impaired is initially assisted by a virtual three-dimensional character that will present the route to guide him to the product. Then, given a query image of an object, the system must display on the screen of the mobile device a list of objects corresponding to the image sent. The problem studied is multidisciplinary and will employ algorithms and methods of object recognition in the area of computational vision combining different descriptors and classifiers for data fusion methods, machine learning approaches for data training, artificial intelligence algorithms and data mining. (AU)