Understanding the Language of the Eye: Detecting and Identifying Eye Events in Real Time via Electrooculography
The human eye has proven to be an enigma; a rich and complex organ that has not quite been fully understood. Naturally, this has caused the eye to be the subject of several medical and anatomical research efforts. More recently, the engineering community has used the eye as the inspiration behind the design and development of cameras, machine vision algorithms, and much more. In order to better understand the eye and its underlying patterns, the technique of electrooculography (EOG) was developed, in which electrical signals originating from the extraocular muscles are measured. The development of EOG has recently opened a new line of research within the engineering disciplines, due to its potential use as an input mechanism for computers and software applications.
In this thesis, I use a pair of Jins Meme glasses, a lightweight and commercially available EOG system, to build a realtime classification system capable of detecting and classifying blinks and movements in the eye. In addition to analyzing the realtime classification system and its inner workings, I use this system to investigate the potential uses of EOG within the realm of Human Computer Interaction. In doing so, I find optimistic results --- the classification system was capable of properly classifying 89.7% of directional eye events and blinks in a set of control datasets. I additionally present two real world use cases for EOG and describe a series of improvements and future work that could aid in advancing the future of EOG.