An Approach for Visual Attention Classification based on Particle Swarm Optimization using Eye Tracker Tests

Abstract

Attention classification has been widely studied over the last decade, with methodologies and proposals for various purposes such as the early detection of autistic spectrum disorder, Attention Deficit Hyperactivity Disorder (ADHD) or just to have a reliable tool to determine whether a subject is being attentive or not. This work proposes a methodology for visual attention classification based on particle swarm optimization using eye tracker data. Firstly, the data was obtained through a series of visual tests applied to a certain number of adult subjects while an eye tracker acquires the eye coordinates. Then, the data was processed to extract the desired features to build the final dataset. To optimize the model, a particle swarm optimization with the K- Means algorithm was performed to generate the optimum groups for classification with KNN. Finally, the performance evaluation and comparison with other works from state of the art were carried out. The proposed methodology reached an accuracy of 97.78\% without using expensive or cumbersome equipment. Therefore, a reliable and comfortable tool for assessing visual attention was achieved using the proposed methodology.