Eeg SUB-BAND spectral centroid frequency and amplitude ratio Features: A comparative study in learning style classification

Abstract

Learning styles are critical element in constructivism that facilitates the process of knowledge creation. Conventional methods to evaluate the psychological trait however are exposed to reliability issues which stem from cultural and language barriers. Hence, a new assessment approach based on the resting EEG is proposed. The paper presents a comparative study between EEG spectral centroid frequency and ratio features in learning style classification. A total of 68 university students have participated in the study. Kolb’s Learning Style Inventory has been implemented to establish the control groups. EEG is then recorded from the antero-frontal region and preprocessed for noise removal. Subsequently, the spectral centroid frequency and amplitude features are extracted. The amplitude component is further normalized via the ratio method. Synthetic EEG is implemented for dataset enhancement. In general, separate investigation via k-nearest neighbor classifier has shown that the spectral centroid frequency outperforms the amplitude ratio components. Alternatively, combination of both features concurrently can effectively improve the overall classification performance.