Comparison of channel selection methods on the classification of EEG data obtained from the animal non-animal categorization experiment

In this study, we have investigated channel selection algorithms on the classification performance of EEG data obtained from animal/non-animal categorization task experiment. Signals from electrodes were analyzed and active locations associated with visual stimuli were determined in the channel selection process. Piecewise Constant Modeling (PCM) and Piecewise Linear Modeling (PLM) techniques were used as feature extraction methods and r (Pearson) values, Fisher Score (FS), Mutual Information (MI), Kullback Leibler Distance (KLD) and Common Spatial Pattern (CSP) methods were used as channel selection methods in the study. It was observed that best classification performance was achieved when PCM was used as feature extraction method and VR was used as channel selection method.