Artefact removal from EEGs using a hybrid BSS-SVM algorithm
L. Shoker, Saeid Sanei · 2004
Various artefacts such as eye-blinking, also called electrooculogram (EOG), and electrocardiograms (ECG) deteriorate the quality of electroencephalograms (EEGs) which are going to be used for clinical diagnosis or to be processed. Blind source separation (BSS) has been a powerful tool to separate different EEG and artefact sources. Identification of different sources from the independent components however, has been always under question. A robust technique for detection and removal of the artefacts requires an efficient classification of the independent components estimated by an effective BSS method. In this talk we will explore different schemes in EEG artefact removal. Then we will concentrate on a robust technique based on fusion of blind identification based BSS algorithm and an effective classification method using support vector machines (SVM). The classifier exploits the statistical properties of both the EEG and the artefact using certain features to identify the best separating hyperplane. We will show that the proposed supervised method is very effective, robust and computationally cost efficient. (19 pages)