Combination of independent component analysis and feature extraction of ERP for level classification of sustained attention

Abstract

This paper investigates the relations between ERP features and visual sustained attention. Continuous Performance Test is used for determining sustained attention level. Fifty eight features were extracted from the 19-channel recorded signals. Twenty four subjects were divided into three classes according to their attention level. LDA classifier is used and high accuracy (94%, 88% and 93% for each two classes) is achieved by using two features in classifying the test data. Obtained results are in agreement with the previous studies.