Comparative Analysis of Feature Selection Techniques for Eye Movement Classification

Electrooculography (EOG) has been widely applied across a range of applications, such as medical diagnosis, assistive technology, and human computer interface interactions. The current study examines the classification of four eye movements (right, left, up, and down) through electrooculography (EOG) signals, testing two feature selection methods: adaptive analysis of variances (ANOVA) and forward feature selection. The objective of this study is to select the most relevant features, among the used statistical, while maintaining a high accuracy level for identifying eye movements. Applying the adaptive ANOVA feature selection method, a maximum classification accuracy of 95.6% was achieved, while the forward feature selection method demonstrated higher performance with an accuracy of 98.7%. The highest accuracy was reached by performing several classification methods in each trial, including K-Nearest Neighbours (KNN), Support Vector Machines (SVM), and Decision Trees (DT). The results indicate the impact of feature selection in enhancing classifier performance.