Single-trial ERP Feature Extraction and Classification for Visual Object Recognition Task

Abstract

This paper describes a classification method of brain signals measured from a subject during visual stimuli presentation. The method is based on exploring the features of visual event related potentials (VERP) in both time and spatial domains. Independent component analysis (ICA) is used for extracting the basic phase-amplitude features. A method for artifact rejection is proposed for excluding the noisy trials. A category-level classifier based on a neural network is trained and its performance on a publicly available database is evaluated and analyzed.