Assistance to Visual Impairing People using Deep Learning and Type-2 Fuzzy Logic

Blind or visually impaired people (BVIP) are increasingly encountering difficulties in their daily lives. The urban congestion of our cities makes it even more difficult for them to get around. Existing assistive devices are not only too expensive, but also not always adapted to our environments. This project proposes a functional approach for the design and implementation of a simple, low-cost, and efficient deep learning-based solution for BVIP. A faster region-based convolutional Neural network will be used for object identification and face recognition, while type-2 fuzzy logic will be implemented for autonomous navigation. The novelty of this approach lies in the presentation of a complete solution integrating deep learning structures in a portable, low-cost, reliable and high-performance hardware. Moreover, the platform will benefit from a high flexibility by allowing the addition of new objects to its internal database in a simple way to meet the changing needs of the user.