Classification of EEG Features Extracted from Classroom Experiment using Weighted K-Nearest Neighbors

Abstract

Electroencephalogram (EEG) modality is one of the most used neuroimaging techniques to uncover the underlying brain waves in different conditions. In recent years, EEG has been used widely in education especially in classroom learning. In this study, 26 participants were selected based on questionnaires to participate in mathematics classroom experiment to detect student’s interest using EEG. A hybrid method combining Empirical Mode Decomposition (EMD) and wavelet transform was developed and employed for feature extraction. The proposed method achieved high classification accuracy using weighted k-Nearest Neighbors (kNN). The high classification accuracy of 85.7% suggests that brain oscillation patterns of high interest students are somewhat different than students with low or no interest.