EEG-based Image Classification using Machine Learning Algorithms

Abstract

This research work focuses on evaluating the most efficient algorithm for EEG-based Image classification by comparing different Machine Learning (ML) algorithms and Ensemble methods. In this analysis, EEG signals are collected with the g.Nautilus headset and analyzed through Independent Component Analysis (ICA) in the EEG Lab. The experiment setup includes six stages, in the first step an image of the object is presented to the user for a few seconds and in the second stage, an image with an object name is presented to the user for the next few seconds. These two steps are repeated for a total of three shapes. Multiple ML algorithms are applied and evaluated to classify visual and written images examined by the user. Moreover, it is concluded that feature engineering, principal component analysis, and parameter tuning are crucial steps in Machine Learning to gain higher performance. The final results prove the capabilities of developing an Image Classification System with K-Nearest Neighbors and Brain-Computer Interface.