Abstract
In this work we will present a method based on singular spectrum analysis (SSA) to remove ocular artifacts from an Electroencephalogram (EEG). After embedding the EEG signals in a feature space of time- delayed coordinates, the principal directions are computed. Using the projections of the embedded signals onto the eigenvectors corresponding to the largest eigenvalues, the Electrooculogram ...
Abstract
In this work we will present a method based on singular spectrum analysis (SSA) to remove ocular artifacts from an Electroencephalogram (EEG). After embedding the EEG signals in a feature space of time- delayed coordinates, the principal directions are computed. Using the projections of the embedded signals onto the eigenvectors corresponding to the largest eigenvalues, the Electrooculogram (EOG) signal is extracted. Thereby it is tacitly assumed that, as the EOG artifact represents a large-amplitude signal, it should be associated with the largest eigenvalues. We incorporate a Minimum Description Length (MDL) criterion based on information theory to determine the appropriate number of eigenvalues which correspond to the EOG signal. The extracted EOG signal is subtracted from the original EEG signal to obtain the artifact- free signal we are interested in.
More literature (via CORE)