Automatic assistant for better mobility and improved cognition of partially sighted persons

In these paper we introduce a novel computer vision assistant for autonomous navigation of partially sighted people. We begin by detecting any type of static and dynamic obstacle present in the scene. Then, we introduce an adapted version of HOG (Histogram of Oriented Gradients) descriptor incorporated into the BoVW (Bag of Visual Words) retrieval framework and demonstrate how this combination can be used for obstacle classification. The design is completed with an acoustic feedback that alert user of potential hazards. The audio bone conduction is employed to allow the visually impaired to hear other sounds from the environment. At the hardware level, the system is totally integrated on a smartphone which makes it easy to wear, non-invasive and low-cost.