Eye Movement Detection using Histogram Oriented Gradient and K-Nearest Neighbors

Abstract

Hand defects will hinder human activities in interacting. In general, people with hand defects will replace the function of the hand with doing an activity in other organs, such as the feet, head, or eyes. Eye movement technologies have been used for different purposes in various industries, from medical, gaming, controlling display menu to assistive technology for people with disabilities. The development of new technologies focused on disability is vital as it improves the quality of life and incorporation into society. Therefore, eye movement detection can later be developed to select menus from the LCD screen. Detection of eye movements is divided into four classes, namely looking up, down, left, and right. The menu on the LCD screen contains options of activities such as calling the nurse, choosing a food menu, and going to the toilet, making it easier for nurses to understand the user's activities. This system is made using Histogram Oriented Gradient combining with Haar Cascade for eye detection and K-Nearest Neighbors as a classifier. Finally, this work shows that the KNN can be used to classify with an average accuracy of 83.33%.