Multi-functional Smart E-Glasses for Vision-Based Indoor Navigation

Wearable sensing and feedback device could provide useful information for the people with visual impairments. In this paper, we develop a real-time, multifunctional electronic eyeglasses (E-Glasses) for vision-based indoor navigation. Our proposed E-glasses has two main functions, which are indoor navigation through navigation algorithm, and object of interest detection through deep learning technology. The safe motion trajectory and the object information are summarized and feedback will be given to the people with visual impairments. The system can run at 39 fps and achieve mean average precision of 84.77% in the test set of images we collected. Detailed experiments of "Navigation" and "Find the target" had an overall success rate above 91%, demonstrating the feasibility of E-Glasses.