Recognition of Indoor Landmarks - an aid to the Visually Impaired

Abstract

Providing sufficient information to visually impaired persons (VIPs) for navigation has always been a tough problem to tackle, and indoor navigation more so since GPS is not available. Thus we must resort to other means. One approach we could take is to install additional hardware, but this is often costly and time-consuming, so we would like to take an alternative approach. Another issue is that the instructions are sometimes quite long, and there is no way that the VIPs can verify that they are on the correct path.

In this project, I attempt to address these issues by creating an iPhone application that helps VIPs have an easier time navigating in MTR stations without GPS. In order for the process of setting up different locations to be as automatic as possible, some machine learning is used to decide the appropriate decision boundary. A feasibility study between several MTR stations is also conducted to determine the accuracy of a few machine learning algorithms in order to decide which one will yield better results.

The final part of the project is verifying that users are taking the correct path. A total of two solutions are proposed.