Indoor human localization based on the corneal reflection of illumination

Corneal imaging has much potential for the development of eye-based interactions. However, it can only provide information on the object being focused on. We therefore propose a localization method based on corneal imaging that exploits the reflections of illumination features from the cornea. A virtual corneal image can be generated from an illumination map, and its similarity to the input eye image can be computed. Global and local localizations are then achieved based on this similarity and a particle filter. The x--, y-- coordinates and θ angle of a participant in a room can thus be estimated practically, as demonstrated experimentally.