Class-adaptive denoising for EEG data classification

Abstract

Brain-computer interface (BCI) systems use electro-encephalogram (EEG) data to control external electronic devices. The main task of BCI systems is to differentiate the classes of mental tasks from the EEG data. The EEG data is inherently complex and difficult to analyze due to interference by eye and muscle movements as well as electrical grid noise. In this paper we analyze shrinkage functions for signal filtering and propose a class-adaptive method for EEG data denoising. The results are evaluated using a Support Vector Machine. © 2012 Springer-Verlag Berlin Heidelberg.