Autoregressive modeling-based feature extraction of EEG/EOG signals

Abstract

Electroencephalogram (EEG) records the electrical patterns of the brain, whereas electrooculogram (EOG) represents the same for eye. Both are complex and are challenged by various artifacts. Therefore, preprocessing and feature extraction of these signals are a vital task so that artifacts can be suppressed and can be used for applications like biometric authentication systems. EOG is present as an artifact in EEG, but combination of EEG and EOG improves the feature recognition and classification accuracy. In this paper, firstly the EEG and EOG signals are preprocessed with the combination of stationary wavelet transform (SWT) and independent component analysis (SWT + ICA) to eliminate some unwanted components from EEG. For EOG signal processing, empirical mode decomposition (EMD) is used to extract eye blinks from the EEG data. After the preprocessing stage, feature extraction is carried out depending on time delineation in case of EOG and autoregressive modeling (AR) technique in case of EEG data. The extracted features of EEG/EOG can be fused together to develop a multimodal system which possesses high classification and recognition accuracy.