Capturing Four Typical Eye Movement Signals Hidden in Electroencephalograph

摘要

We proposed a process to find and recognize four typical eye movement signals hidden in EEGs. Eye movement signals contained in EEG are usually looked as artefacts by doctors. But they can be useful if the movements are intentionally launched for special purpose. Identifying typical eye movements from a flow of EEG signals can facilitate communications. We introduce the way to identify four typical eye movement signals by analyzing the electrooculography (EOG) signals with kernel principal angles. Supposing one person tries to give signal by moving eyes, it is convenient for him or her to move the eyes left, right, up or down. We sampled many typical EOG signals of these special conditions. Kernel principal angles are often used to measure the similarity of two data sets. With the help of kernel principal angles, we tried to capture the eye movement signals hidden in EEG signals and recognize them. The final experiments show that overwhelming majority of the cosines in the same classes are over 0.95, and cosines over different classes are less than 0.80. This means that kernel principal angles can be effective to capture and identify typical eye movement signals.