A template-based isomap algorithm for real-time removal of ocular artifacts from EEG signals

Abstract

Frequent occurrences of ocular artifacts seriously interfere with the electroencephalogram (EEG) interpretation and analysis. In present study, a novel technique is presented to eliminate ocular artifacts from EEG signals in real-time. The usual operation mode of manifold learning algorithm is modified to improve the effect of ocular artifacts removal. Topographic and spectral features were extracted from the components decomposed by independent component analysis (ICA) from the EEG signals. Then a template-based Isometric Mapping algorithm is proposed to reduce dimensionality of these original features. At last, the samples with low-dimensional features are fed to a classifier, which is selected from several typical classifiers by comparing their classification performances, to identify ocular artifacts components. The classification results show that the template-based Isometric Mapping algorithm with the nearest neighbor classifier performs well. The averaged effects of ocular artifacts removal show that the novel technique can effectively remove ocular artifacts in the real-time application.